System

Abstract

A system is provided.SOLUTION: A system comprising: means for collecting transaction data and delivery destination information of a credit card; means for normalizing the transaction data and the delivery destination information, removing abnormal values, and extracting feature values; means for training a generative artificial intelligence model using the normalized and feature-value-extracted data; means for receiving transaction data in real time and detecting an abnormality using the generative artificial intelligence model; and means for notifying a user of the detected abnormality and allowing the user to confirm the abnormality.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] In recent years, there have been frequent incidents of fraudulent credit card use and fraudulent changes to delivery addresses, and the damage caused by these incidents is expanding. When fraud occurs, it places a heavy emotional and financial burden on the victim, and detecting and responding to it requires significant resources. Conventional systems have struggled to detect fraud in real time, making it difficult to respond quickly. Therefore, there is a need for the development of a system that can quickly and accurately detect fraudulent use in credit card transactions and product delivery, and quickly notify users. [Means for solving the problem]

[0005] The present invention includes a means for collecting and normalizing credit card transaction data and shipping address information, removing outliers, and extracting features. It also includes a means for using a generative artificial intelligence model to learn normal usage patterns and the characteristics of fraudulent usage, and for receiving and analyzing transaction data in real time to detect anomalies. If fraudulent usage is detected, the system also includes a means for promptly notifying the user and allowing the user to confirm the anomaly. If the user determines that fraudulent usage has occurred, the system provides a system that enables prompt responses, such as suspending the card or changing the shipping address. It also includes a means for periodically updating the transaction data and shipping address information and re-learning the model to improve its accuracy.

[0006] "Credit card transaction data" refers to all data related to transactions conducted using a credit card, including, specifically, the transaction amount, transaction date and time, transaction location, transaction details, etc.

[0007] "Shipping information" refers to detailed information such as the address and contact details of the destination to which goods or services will be delivered.

[0008] "Normalization" refers to the process of converting collected data into a consistent format to maintain data quality and consistency.

[0009] "Outlier removal" refers to the process of removing data points that fall outside the normal range, and is done to improve the accuracy of data analysis.

[0010] "Feature extraction" refers to the process of extracting important data points or attributes from raw data for use in analysis.

[0011] A "generative artificial intelligence model" refers to an artificial intelligence model that has the ability to learn patterns based on input data and generate new data.

[0012] "Normal usage patterns" refer to data patterns associated with the normal use of credit cards and delivery of goods.

[0013] "Fraud signatures" refer to characteristics or trends in data that indicate unusual patterns or fraudulent activity.

[0014] "Means of receiving and analyzing transaction data in real time" refers to a mechanism that instantly sends data to a server the moment a user makes a transaction and automatically analyzes it.

[0015] "Anomaly detection measures" refers to mechanisms for identifying and confirming the existence of data patterns that deviate from normal ranges.

[0016] "Means of notifying users" refers to the means of quickly conveying information to users when an abnormality is detected, and specifically includes email, SMS, push notification, etc.

[0017] "A system that allows for quick responses such as suspending cards and changing delivery addresses" refers to a mechanism that allows a user to immediately suspend credit card use or change the delivery address for products if they confirm fraudulent use.

[0018] "Retraining means" refers to the process of retraining a generative AI model on new data to maintain or improve its performance and accuracy. [Brief explanation of the drawings]

[0019] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5]FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0020] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0021] First, the terms used in the following description will be explained.

[0022] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0023] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0024] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0025] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0026] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0027] [First embodiment]

[0028] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0029] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0030] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0031] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0032] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0033] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0034] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0035] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0036] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0037] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0038] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0039] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0040] This invention is a system that collects credit card transaction data and shipping destination information, and uses a generative artificial intelligence model to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[0041] Data collection and preprocessing

[0042] The server collects credit card transaction data and shipping information for each transaction, as well as known fraud data. The collected data is normalized and outliers are removed. This extracts features and prepares the data for processing.

[0043] Examples:

[0044] When a user purchases a product from an online shop, transaction data (purchase date and time, amount, shop location, etc.) is collected. At the same time, delivery address information for the product is also obtained.

[0045] Model training and generation

[0046] The server uses the preprocessed data to train a generative artificial intelligence model, which learns normal usage patterns and the characteristics of fraudulent usage, and has the ability to detect anomalies with high accuracy.

[0047] Examples:

[0048] The server analyzes past transaction data to learn, for example, which region a user typically spends money in. This allows the model to memorize typical usage patterns.

[0049] Real-time detection

[0050] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, an alert is generated immediately.

[0051] Examples:

[0052] If a user makes a sudden, expensive purchase in a particular major city while traveling abroad, the transaction data is sent to the server, and the model determines that it differs from the normal pattern, detecting the anomaly and generating a warning.

[0053] User notification and response

[0054] If the server detects an abnormality, it will immediately notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user will take steps to suspend the card or change the shipping address.

[0055] Examples:

[0056] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and, for example, determines that the transaction was not initiated by them, they can press a button in the app to immediately report the incident to their credit card company and take steps to suspend their card.

[0057] Continuous model updates

[0058] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy, allowing the system to always perform highly accurate fraud detection based on the latest data.

[0059] Examples:

[0060] Through monthly data updates, the server learns new fraud patterns and retrains the model, keeping the system up to date with the latest fraud techniques.

[0061] In this way, the system of the present invention detects fraudulent credit card use and fraudulent changes to shipping addresses in real time with high accuracy, and provides prompt notification and response to users, thereby realizing a safe and secure transaction environment for users.

[0062] The processing flow will be explained below.

[0063] Step 1:

[0064] The server periodically collects credit card transaction data and shipping information.

[0065] Specific operation: Using the APIs of credit card companies and delivery companies, the system obtains the latest transaction data and delivery address information, while also collecting data on known past fraudulent use.

[0066] Step 2:

[0067] The server normalizes the collected data, removes outliers, and extracts features.

[0068] Specific operations: Extracts key fields such as date, time, amount, store used, and delivery destination from raw data and normalizes them into a consistent format. Detects and removes outliers and calculates the features required for analysis.

[0069] Step 3:

[0070] The server trains a generative artificial intelligence model using the preprocessed data.

[0071] Specific operations: Split the normalized and feature-extracted data into a training set and a test set, and train a generative AI model to learn normal usage patterns and the characteristics of fraudulent usage.

[0072] Step 4:

[0073] When a user makes a transaction using a credit card, the terminal transmits the transaction data to the server in real time.

[0074] What it does: The POS system or online payment platform instantly sends the user's transaction data (e.g., time, location, amount, etc.) to the server.

[0075] Step 5:

[0076] The server immediately analyzes the received transaction data using a generative artificial intelligence model and detects any anomalies.

[0077] How it works: Incoming data is fed into a generative AI model in real time, where it is compared with normal usage patterns. If an abnormal pattern is detected, an anomaly flag is raised.

[0078] Step 6:

[0079] If an abnormality is detected, the server will promptly notify the user.

[0080] Specific behavior: If an anomaly is flagged, a warning message will be sent using the user's contact information (email address, SMS, push notification, etc.).

[0081] Step 7:

[0082] The user receives a notification and verifies the authenticity of the transaction.

[0083] Specific operation: When a user receives a notification, they access a web portal or smartphone app to verify whether the transaction was initiated by them.

[0084] Step 8:

[0085] If the user determines that the use has been fraudulent, procedures such as suspending the card or changing the delivery address will be implemented.

[0086] Specific actions: If a user determines that a transaction has been fraudulent, they can contact their credit card company using a link in the dedicated app or web portal to suspend their card. They can also contact the delivery company to suspend or change the delivery of their product.

[0087] Step 9:

[0088] The server periodically updates transaction data and delivery destination information and retrains the generative artificial intelligence model.

[0089] What it does: Periodically prepare new datasets to evaluate the performance of existing generative AI models, retraining them as needed to maintain or improve their accuracy.

[0090] Through the above steps, the system of the present invention detects credit card fraud and shipping address fraud in real time, promptly notifying the user so that appropriate action can be taken.

[0091] Example 1

[0092] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0093] Fraudulent credit card use can cause serious financial damage to users. Early detection of fraudulent use and appropriate countermeasures are required, but current systems have difficulty detecting fraud in real time with high accuracy. Furthermore, maintaining the accuracy of the system requires continuous model updates, which are difficult to achieve efficiently with conventional methods.

[0094] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0095] In this invention, the server includes means for collecting credit card transaction data and recipient information, means for normalizing the transaction data and recipient information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, means for notifying the user of detected anomalies, allowing the user to confirm the anomaly and, if fraudulent, suspending the payment method or changing the recipient information, and means for periodically updating the transaction data and recipient information and relearning the generative artificial intelligence model to maintain or improve its accuracy. This makes it possible to detect fraudulent credit card use in real time with high accuracy and provide users with prompt notification and support.

[0096] A "credit card" is a payment method issued by a bank or credit card company that users use when purchasing goods or services.

[0097] "Transaction data" refers to information relating to purchases or payments made using a credit card, including, for example, the date and time of purchase, the amount, and the place of purchase.

[0098] "Recipient information" refers to information about the delivery address of purchased products or services, and is data including information such as the address, recipient name, and contact details.

[0099] "Normalization" is the process of converting data values ​​to a uniform scale to ensure data consistency and comparability.

[0100] An "outlier" is a value that deviates from the normal data pattern and should be treated as noise or an error in data analysis.

[0101] In data analysis and machine learning, a "feature" is a numerical value or indicator that represents an important attribute or pattern of data.

[0102] A "generative artificial intelligence model" is a type of artificial intelligence that is an algorithm that learns patterns and characteristics of data and makes predictions and detects anomalies in new data.

[0103] "Real-time" refers to processing data either instantly as it occurs or with very little delay.

[0104] "Anomaly detection" is the process of identifying data that deviates from normal patterns, indicating potential problems or fraud.

[0105] "User" means an end user who purchases goods or services using a credit card and receives notifications from the system.

[0106] "Payment Instrument" means a method of paying for goods and services, including credit cards and debit cards.

[0107] "Model retraining" is the process of using new collected data to readjust the model parameters and algorithms in order to maintain or improve the accuracy of a generative artificial intelligence model.

[0108] This invention is a system that collects credit card transaction data and recipient information, and uses a generative artificial intelligence model to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[0109] Data collection and preprocessing

[0110] The server collects credit card transaction data and payee information for each transaction, as well as known fraud data. Through an API, the server normalizes the collected data and removes outliers. This extracts features and prepares the data for processing.

[0111] Examples:

[0112] For example, when a user purchases a product from an online shop, the transaction data (purchase date and time, amount, shop location, etc.) is sent to the server. At the same time, the server also collects information about the recipient of the product.

[0113] Model learning

[0114] The server uses the preprocessed data to train a generative AI model, which then learns normal usage patterns and the characteristics of fraudulent usage, detecting anomalies with high accuracy.

[0115] Examples:

[0116] The server analyzes past transaction data to learn, for example, which region a user typically spends money in. This allows the server to memorize typical usage patterns into a model.

[0117] Real-time detection

[0118] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, an alert is generated immediately.

[0119] Examples:

[0120] For example, if a user makes a large purchase in a particular major city while traveling abroad, the transaction data is sent to the server, and the model determines that the transaction differs from the normal pattern. As a result, an anomaly is detected and an alert is generated.

[0121] User notification and response

[0122] If the server detects an abnormality, it will immediately notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user can take steps to stop the payment method or change the payee information.

[0123] Examples:

[0124] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and, for example, determines that the transaction was not made by them, they can press a button in the app to immediately report it to their credit card company and take steps to suspend their credit card.

[0125] Continuous model updates

[0126] The server periodically updates transaction data and recipient information, and retrains the generative AI model to maintain or improve its accuracy, allowing the system to always perform highly accurate fraud detection based on the latest data.

[0127] Examples:

[0128] Every month, the server retrains the model using a new dataset collected, updating the model parameters to the latest state, thereby keeping the system adaptable to new fraud techniques.

[0129] Prompt Sentence Examples

[0130] "Please retrain your fraud detection model using this month's transaction data."

[0131] This system makes it possible to detect fraudulent credit card use in real time with high accuracy, and to provide users with prompt notification and response.

[0132] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0133] Step 1:

[0134] Data collection

[0135] The server collects transaction data and recipient information from the APIs of credit card companies and online shops. Inputs include the purchase date and time, amount, shop location, and recipient information for each transaction. This data is stored on the server.

[0136] Specific behavior:

[0137] When a new transaction occurs, the server receives the transaction data in real time through the API. For example, when a user purchases a product, the transaction information is sent to the server.

[0138] Step 2:

[0139] Fraud data collection

[0140] The server also collects known fraud data. The input is transaction data of previously detected fraudulent activity. This data serves as the basis for learning fraud patterns.

[0141] Specific behavior:

[0142] The server periodically downloads fraud data from credit card companies and other data providers.

[0143] Step 3:

[0144] Data normalization and outlier removal

[0145] The server normalizes the collected transaction data and recipient information and removes outliers. The input is the collected raw data, and the output is the normalized data. This keeps the data consistent and makes it easier to analyze.

[0146] Specific behavior:

[0147] Check the transaction dataset stored in the database and perform preprocessing such as scaling and missing value imputation.

[0148] Step 4:

[0149] Feature extraction

[0150] The server extracts features from the preprocessed data. The input is normalized data, and the output is a set of features. The features include transaction frequency, amount statistics, geographic information, etc.

[0151] Specific behavior:

[0152] For each transaction, multiple metrics (purchase date and time, average and variance of the amount, geographical distance, etc.) are calculated to create a feature set.

[0153] Step 5:

[0154] Learning generative artificial intelligence models

[0155] The server uses the extracted features to train a generative AI model. The input is a set of features, and the output is a trained generative AI model. This allows the server to learn the characteristics of normal usage patterns and fraudulent usage.

[0156] Specific behavior:

[0157] Historical trading data is used as a training set to optimize the model parameters.

[0158] Step 6:

[0159] Real-time data transmission

[0160] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The input is the user's transaction data, and the transmitted data arrives at the server.

[0161] Specific behavior:

[0162] The user's purchase information is immediately sent from the terminal to the server.

[0163] Step 7:

[0164] Real-time analytics

[0165] The server instantly analyzes the received transaction data using a generative artificial intelligence model. The input is the transaction data sent in real time, and the output is the presence or absence of anomalies. If there is a possibility of fraud, a warning is generated.

[0166] Specific behavior:

[0167] The data is analyzed layer by layer, and when an anomaly is detected, an anomaly score is generated.

[0168] Step 8:

[0169] User Notifications

[0170] If an abnormality is detected, the server will promptly send a notification to the user. The input is the abnormality detection result, and the output is a notification message to the user.

[0171] Specific behavior:

[0172] When an anomaly is detected, the server generates a notification message and sends a push notification to the user's smartphone.

[0173] Step 9:

[0174] User Support

[0175] The user receives a notification and verifies whether the transaction is legitimate. If it is fraudulent, they can block the payment method or change the payee information. The input is the user's confirmation, and the output is an action such as reporting to the card company.

[0176] Specific behavior:

[0177] If the user checks the notification and presses the "This transaction was not made by me" button, the transaction will be reported to the credit card company and the credit card will be suspended.

[0178] Step 10:

[0179] Model Update

[0180] The server periodically updates the transaction data and recipient information and retrains the generative AI model to maintain or improve its accuracy. The input is a new dataset and the output is an updated generative AI model.

[0181] Specific behavior:

[0182] Every month, the model is retrained using the latest collected data and the model parameters are optimized.

[0183] (Application example 1)

[0184] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0185] Conventional credit card fraud detection systems were slow to detect fraud, with users often only realizing they had been compromised after the fraud had actually occurred. Furthermore, they lacked the functionality to allow users to respond quickly after fraud was detected, making it difficult to prevent fraudulent use. Furthermore, as fraud patterns change, continuous system updates are required, which was difficult to achieve with conventional systems.

[0186] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0187] In this invention, the server includes means for collecting credit card transaction data and shipping destination information, means for normalizing the transaction data and shipping destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for receiving the transaction data in real time and detecting anomalies using the generative artificial intelligence model, means for sending a push notification of the detected anomaly to the user's communication terminal, and means for allowing the user to immediately take steps to suspend their credit card based on the results of the generative artificial intelligence model. This enables highly accurate detection of fraudulent use in real time and provides an environment in which users can respond quickly.

[0188] A "credit card" is a payment method with a certain credit limit that is used when purchasing goods and services.

[0189] "Transaction Data" refers to detailed information about purchases made using a credit card, including transaction date and time, amount, store location, and other data.

[0190] "Shipping destination information" refers to information about the shipping destination of products purchased with a credit card, and includes the address and name of the recipient.

[0191] "Normalization" is the process of converting data into a consistent format, a technique that makes data analysis and model training easier by standardizing the scale of variables.

[0192] An "outlier" is a piece of data that is statistically significantly different from other data points, and is a number that may affect the results of an analysis.

[0193] "Features" are input data provided to a machine learning model that represent characteristics or attributes that are important for the model to learn data patterns.

[0194] A "generative artificial intelligence model" is a machine learning algorithm designed to learn the patterns and rules needed to perform a specific task, and is typically powered by large amounts of training data.

[0195] "Learning" is the process by which a generative artificial intelligence model finds patterns and rules in given data and improves its performance.

[0196] "Real-time" refers to a method in which data is processed as it is acquired and results are provided immediately.

[0197] "Anomaly detection" is a function that finds behaviors or values ​​in data that differ from normal patterns, and is used to identify fraudulent use or abnormal behavior.

[0198] A "push notification" is a notification message that an application sends directly to a user's communication terminal, and is a means of delivering information to the user immediately.

[0199] "Credit card suspension procedure" is a process to temporarily or permanently disable the use of a credit card suspected of fraudulent use, and is carried out to prevent further damage.

[0200] The system for realizing this invention collects credit card transaction data and shipping destination information, detects fraudulent use in real time using a generative artificial intelligence model, and notifies users. This system is composed of a server, terminals, and users.

[0201] Data collection and preprocessing

[0202] The server collects credit card transaction data and shipping address information for each transaction. The collected data is normalized and outliers are removed. This allows features to be extracted and the data is ready for processing. For example, when a user purchases a product online, the transaction data (purchase date and time, amount, purchase location, etc.) is collected. At the same time, the product's shipping address information is also obtained.

[0203] Model learning

[0204] The server uses the preprocessed data to train a generative artificial intelligence model. This model learns the characteristics of normal usage patterns and fraudulent usage, and has the ability to detect anomalies with high accuracy. For example, the server analyzes past transaction data to learn which region a user is shopping in and within what price range. This allows normal usage patterns to be memorized in the model.

[0205] Real-time detection

[0206] When a user purchases a product, the user's device sends the transaction data to a server in real time. The data received by the server is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, a warning is immediately generated. For example, if a user makes an expensive purchase in an unusual location while traveling abroad, the transaction data is sent to the server and is determined to be abnormal by the model. As a result, the anomaly is immediately detected and a warning is generated.

[0207] User notification and response

[0208] If an abnormality is detected, the server will promptly send a notification to the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user can take steps to suspend their credit card. For example, when the server detects an abnormality, a push notification will be sent to the user's smartphone. If the user checks the notification and determines that the transaction was not made by them, they can press a button in the app to report it to their credit card company and take steps to suspend their card.

[0209] Continuous model updates

[0210] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy. This allows the system to always perform highly accurate fraud detection based on the latest data. For example, the server updates data monthly, learns new fraudulent usage patterns, and retrains the model. This makes it possible to respond to the latest fraud techniques.

[0211] Prompt Sentence Examples

[0212] Examples of prompts that can be provided to generative AI models include:

[0213] "Does the purchase of over 5,000 yen in Shinjuku Ward match a pattern that has not been identified as abnormal in your past transaction history?"

[0214] The above is a specific embodiment of the present invention. This system makes it possible to detect fraudulent use with high accuracy in real time and to provide an environment in which users can respond quickly.

[0215] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0216] Step 1:

[0217] The server collects credit card transaction data and shipping information. Specifically, it obtains detailed information about each transaction (transaction date and time, amount, store location, shipping information, etc.) through a data collection API. The input is the detailed information about each transaction, and the output is a list of the collected transaction data and shipping information.

[0218] Step 2:

[0219] The server normalizes the collected transaction data and shipping destination information and removes outliers. Specifically, it performs a data cleansing process to remove outliers and missing values ​​and format the data into a unified format. The input is a list of collected transaction data, and the output is the normalized and cleansed data.

[0220] Step 3:

[0221] The server extracts features from the normalized and cleansed data. Specifically, it calculates and extracts features for each transaction by amount, date, time, and region. The input is the normalized and cleansed data, and the output is the data with extracted features.

[0222] Step 4:

[0223] The server trains a generative AI model using the data from which features have been extracted. Specifically, it performs model training to learn normal transaction patterns and fraudulent usage patterns based on past transaction data. The input is the data from which features have been extracted, and the output is the trained generative AI model.

[0224] Step 5:

[0225] When a new transaction occurs, the user's device sends the transaction data to the server in real time. Specifically, the data transmission API is called the moment the transaction is completed. The input is the new transaction data, and the output is the transaction data sent to the server.

[0226] Step 6:

[0227] The server analyzes the transaction data received in real time using a generative artificial intelligence model to detect anomalies. Specifically, the transaction data is input into the model, and the results are judged to be abnormal. The input is the transaction data received in real time and the trained model, and the output is the judgment result of whether or not there is an abnormality.

[0228] Step 7:

[0229] If an anomaly is detected, the server promptly sends a push notification to the user's device. Specifically, it uses the notification service API to send the details of the anomaly detection to the user's device. The input is the detected anomaly information, and the output is the push notification sent to the user.

[0230] Step 8:

[0231] The user receives a push notification on their device and verifies the legitimacy of the transaction. If they determine that the transaction is fraudulent, they can proceed with suspending their credit card through the application. Specifically, by pressing a button within the application, fraudulent use is reported and the card is suspended. The input is the user's confirmation result, and the output is the completion of the card suspension procedure.

[0232] Step 9:

[0233] The server periodically updates transaction data and delivery destination information, and retrains the generative AI model to improve its accuracy. Specifically, it periodically updates data and trains it to learn new fraudulent usage patterns. The input is the latest transaction data, and the output is the updated generative AI model.

[0234] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0235] This invention is a system that collects credit card transaction data and shipping destination information, and uses a generative AI model combined with an emotion engine to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[0236] Data collection and preprocessing

[0237] The server collects credit card transaction data and shipping information for each transaction, as well as known fraud data. The collected data is normalized and outliers are removed. This extracts features and prepares the data for processing.

[0238] Examples:

[0239] When a user purchases a product from an online shop, transaction data (purchase date and time, amount, shop location, etc.) is collected. At the same time, delivery address information for the product is also obtained.

[0240] Model training and generation

[0241] The server uses the preprocessed data to train a generative AI model. This model learns normal usage patterns and the characteristics of fraudulent usage, and is capable of detecting anomalies with high accuracy. Meanwhile, the emotion engine recognizes emotions from the user's voice, facial expressions, and text input, and analyzes the data. Emotional data is also used to train the generative AI model.

[0242] Examples:

[0243] The server analyzes past transaction data to learn, for example, which region a user typically shopped in and within what price range. This allows the model to memorize typical usage patterns. At the same time, it also analyzes user emotional data and incorporates it into the model.

[0244] Real-time detection

[0245] When a user purchases a product, the device sends the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an abnormality is detected, an alert is generated immediately. The user's reaction is also analyzed by an emotion engine, and the user's level of stress and anxiety is also detected.

[0246] Examples:

[0247] If a user makes a sudden, expensive purchase in a specific major city while traveling abroad, the transaction data is sent to the server, and the model determines that it differs from normal patterns. As a result, an anomaly is detected, and analysis by the emotion engine detects the user's feelings of surprise or anxiety.

[0248] User notification and response

[0249] If an abnormality is detected, the server will promptly notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user will go through the necessary procedures to suspend the card or change the shipping address. The emotion engine analyzes the user's reaction and customizes the notification content and response procedures as needed.

[0250] Examples:

[0251] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and identifies that the transaction was not initiated by them, the emotion engine detects that the user is extremely anxious and provides gentle instructions on how to respond. The user can also press a button in the app to immediately report the incident to their credit card company and take steps to suspend their card.

[0252] Continuous model updates

[0253] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy. This allows the system to always perform highly accurate fraud detection based on the latest data. Furthermore, emotion data from an emotion engine is added, enabling more personalized responses.

[0254] Examples:

[0255] The server learns new fraud patterns through monthly data updates and retrains the model, ensuring the system can respond to the latest fraud techniques. Data from the emotion engine is also incorporated into the model, making it possible to customize notification methods and countermeasures based on the user's emotional state.

[0256] In this way, the system of the present invention detects fraudulent credit card use and fraudulent changes to shipping addresses in real time with high accuracy, and provides prompt notification and response that takes into account the user's emotional state, thereby realizing a safe and secure transaction environment for users and reducing their mental burden.

[0257] The processing flow will be explained below.

[0258] Step 1:

[0259] The server periodically collects credit card transaction data and shipping information.

[0260] Specific operation: Using the APIs of credit card companies and delivery companies, the system obtains the latest transaction data and delivery address information, while also collecting data on known past fraudulent use.

[0261] Step 2:

[0262] The server normalizes the collected data, removes outliers, and extracts features.

[0263] Specific operations: Extract key fields such as date, time, amount, store used, and delivery destination from raw data, normalize them into a consistent format, detect and remove outliers, and calculate the features required for analysis.

[0264] Step 3:

[0265] The server trains a generative artificial intelligence model using the preprocessed data.

[0266] Specific operation: The normalized and feature-extracted data is split into a training set and a test set. A generative AI model is trained to learn normal usage patterns and the characteristics of fraudulent usage.

[0267] Step 4:

[0268] When a user makes a transaction using a credit card, the terminal transmits the transaction data to the server in real time.

[0269] Specific operation: The POS system or online payment platform immediately sends the user's transaction data (e.g., time, location, amount, etc.) to the server.

[0270] Step 5:

[0271] The server immediately analyzes the received transaction data using a generative artificial intelligence model and detects any anomalies.

[0272] How it works: Incoming data is fed into a generative AI model in real time, where it is compared with normal usage patterns. If an abnormal pattern is detected, an anomaly flag is raised.

[0273] Step 6:

[0274] If an abnormality is detected, the server will promptly notify the user.

[0275] Specific behavior: If an abnormality flag is raised, a warning message will be sent using the user's contact information (email address, SMS, push notification, etc.).

[0276] Step 7:

[0277] The emotion engine analyzes the user's reaction in real time when they receive a notification and detects their emotions (e.g., surprise, anxiety, anger, etc.).

[0278] Specific operation: The emotion engine analyzes the user's voice, facial expressions, text input, etc. when receiving the notification to determine their emotional state.

[0279] Step 8:

[0280] The user receives a notification and verifies the authenticity of the transaction.

[0281] Specific operation: Upon receiving the notification, the user accesses the web portal or smartphone app to verify whether the transaction was initiated by them.

[0282] Step 9:

[0283] The server customizes the notification content and response procedures based on the user's emotions detected by the emotion engine.

[0284] Specific behavior: If the user's emotions indicate anxiety or surprise, the server will re-notify them with softer language and reassuring messages, and if necessary, instruct them to escalate the issue to a support staff member.

[0285] Step 10:

[0286] If the user determines that the use has been fraudulent, procedures such as suspending the card or changing the delivery address will be implemented.

[0287] Specific operation: If a user determines that a transaction has been fraudulent, they can contact their credit card company using a dedicated app or a link in the web portal to suspend their card. They can also contact the delivery company to suspend or change the delivery of their product.

[0288] Step 11:

[0289] The server periodically updates transaction data and delivery destination information and retrains the generative artificial intelligence model.

[0290] How it works: Regularly prepare new datasets and evaluate the performance of existing generative AI models. Retrain as needed to maintain or improve model accuracy. Additionally, by adding data from the emotion engine, the model can be trained to provide more personalized responses.

[0291] Through these steps, the system of the present invention detects credit card fraud and shipping fraud in real time and enables quick and appropriate responses based on the user's emotional state.

[0292] Example 2

[0293] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0294] In modern society, credit card fraud is on the rise, with the risk increasing especially in online shopping and remote transactions. When fraud occurs, it not only causes financial loss to users, but also causes significant psychological stress. Furthermore, existing fraud detection systems are often inadequate because they struggle to respond in real time and are unable to take into account the user's emotional state. This makes it difficult for many users to conduct transactions with confidence. Therefore, there is a need for a system that can detect fraud quickly and accurately, and provide notifications and responses that take the user's emotions into account.

[0295] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for collecting credit card transaction data and delivery destination information, means for normalizing the transaction data and delivery destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for collecting and analyzing user emotion data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, and means for notifying the user of detected anomalies and allowing the user to confirm the anomaly. This makes it possible to quickly and accurately detect fraudulent credit card use and to notify and respond in consideration of the user's emotional state.

[0296] "Credit card transaction data" means information that records purchases and payments made using a credit card.

[0297] "Delivery destination information" refers to information such as the address and contact details of the recipient of the purchased product or service.

[0298] "Normalization" is the process of standardizing the range and units of data to make it easier to compare and analyze.

[0299] "Outlier removal" is the process of identifying and eliminating invalid data points within a data set.

[0300] "Features" are important input variables used in machine learning models for prediction and classification.

[0301] A "generative artificial intelligence model" is an AI model that has the ability to learn from data, recognize patterns, and generate or analyze new data.

[0302] "User emotion data" refers to information about the user's emotional state obtained from the user's facial expression, voice, text input, and the like.

[0303] "Receiving transaction data in real time" means that the data is sent to the server and received immediately the moment a transaction occurs.

[0304] "Anomaly detection" refers to identifying fraudulent or unusual behavior that deviates from normal patterns.

[0305] "Notification" is the act of the system informing the user of abnormalities or important information.

[0306] This invention relates to a highly accurate and rapid detection system for preventing fraudulent use of credit cards. This system collects credit card transaction data and shipping destination information, and uses this data to learn and apply a generative artificial intelligence model that detects fraudulent use. The details are as follows:

[0307] Data collection and preprocessing

[0308] The server collects credit card transaction data and shipping information for each transaction, which involves retrieving the data through an API and storing it in a database using MySQL, a popular relational database management system (RDBMS).

[0309] Specifically, when a user purchases an item from an online shop, the transaction data (purchase date and time, amount, store location, etc.) and shipping information are stored in a MySQL database. The collected data is read using the Python Pandas library, normalized, and outliers are removed, preparing the data for processing.

[0310] Model training and generation

[0311] The server uses the preprocessed data to train a generative artificial intelligence model. This model learns normal usage patterns and the characteristics of fraudulent usage, and detects anomalies with high accuracy. The specific software used is Scikit-learn's RandomForestClassifier.

[0312] The model training uses transaction data and fraudulent use data collected to date. Additionally, Affectiva's SDK is used as a sentiment analysis library to collect and analyze user sentiment data. This sentiment data is also fed back into the model training.

[0313] Specifically, the server analyzes past transaction data to learn, for example, which region a user typically spends money in. At the same time, it analyzes emotional data using Affectiva's SDK and incorporates that data into the model.

[0314] Real-time detection

[0315] When a user purchases a product, the device sends the transaction data to the server in real time. The server immediately analyzes the received data using a generative artificial intelligence model and immediately generates a warning if an abnormality is detected. The server also analyzes the user's reactions using an emotion engine to detect the user's level of stress and anxiety.

[0316] Specifically, if a user makes a sudden, expensive purchase in a specific major city while traveling abroad, the transaction data is sent to the server, and the model determines that the transaction is different from the usual pattern. If it determines that fraudulent use is suspected, emotion analysis can also detect the user's anxiety.

[0317] User notification and response

[0318] The server uses Firebase Cloud Messaging as a communication library to promptly notify the user if an abnormality is detected. The user who receives the notification can check whether the transaction is legitimate, and if it is fraudulent, can immediately take action such as suspending the card or changing the delivery address.

[0319] Specifically, when the server detects an abnormality, it sends a push notification to the user's smartphone. If the user checks the notification and determines that the transaction was not initiated by them, the app uses emotional analysis to detect that the user is extremely anxious and provides gentle guidance. The app also allows the user to press a button to report the transaction to their credit card company and immediately take steps to suspend their card.

[0320] Continuous model updates

[0321] The server periodically updates transaction data and delivery destination information, and performs re-training to improve the accuracy of the generative AI model. This ensures highly accurate fraud detection based on the latest data. Emotional data is also added, enabling more personalized responses.

[0322] Specifically, the server updates data monthly, learns new fraudulent usage patterns, and retrains the model. This ensures the system can respond to the latest fraud techniques. Emotional data is also fed back to the model, allowing it to provide notification methods and countermeasures tailored to the user.

[0323] Prompt Sentence Examples

[0324] Fraud detection is performed on newly collected transaction data, including:

[0325] Purchase date: 2023-10-01 10:00:00

[0326] Amount: $500

[0327] Shop location: London

[0328] Delivery Address: 123 ABC Street, London

[0329] User emotion data: Moderate anxiety

[0330] Use this data to detect anomalies in real time using fraud detection models and notify users as needed.

[0331] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0332] System program processing flow

[0333] Step 1:

[0334] The server collects credit card transaction data and shipping information. Specifically, it uses an API to retrieve credit card transaction information and stores it in a MySQL database. The input includes the transaction data and shipping information retrieved from the API. The output is this raw data stored in the database.

[0335] Specific behavior:

[0336] Make API calls to retrieve transaction data and shipping information.

[0337] The retrieved data is stored in a MySQL database.

[0338] Step 2:

[0339] The server pre-processes the collected data. It uses the Pandas library to normalize the data, remove outliers, and extract features. The input includes raw data stored in a MySQL database. The output is the pre-processed, feature-extracted data.

[0340] Specific behavior:

[0341] Use Pandas to read data from a MySQL database.

[0342] Normalize the data to make it consistent across ranges.

[0343] Detect and remove outliers.

[0344] Important features (e.g., purchase date and time, amount, store location) are extracted and saved as structured data.

[0345] Step 3:

[0346] The server uses the preprocessed data to train a generative AI model. It uses Scikit-learn's RandomForestClassifier to learn normal transaction patterns and fraudulent usage patterns. The input includes training data with extracted features. The output is a trained generative AI model.

[0347] Specific behavior:

[0348] Define a model using Scikit-learn's RandomForestClassifier.

[0349] The data from which the features have been extracted is supplied to the model as training data and fitted.

[0350] The model is evaluated and parameters are adjusted as necessary.

[0351] Step 4:

[0352] The server collects and analyzes the user's emotional data. Using Affectiva's SDK, it analyzes the user's emotions (e.g., stress, anxiety, surprise, etc.) and stores the data. Inputs include the user's facial expressions, voice, and text input. The output is the analyzed emotional data.

[0353] Specific behavior:

[0354] Affectiva's SDK is used to analyze the user's facial expressions, voice, and text input.

[0355] The analysis results (emotion data) are saved as structured data and provided to a generative artificial intelligence model.

[0356] Step 5:

[0357] When a user purchases a product, the device sends transaction data to a server in real time. The server analyzes the received data using a generative artificial intelligence model to detect anomalies. The input includes real-time transaction data and a pre-trained model. The output is a warning message or notification when an anomaly is detected.

[0358] Specific behavior:

[0359] The terminal collects transaction data and immediately transmits it to the server.

[0360] The server receives the transaction data and inputs it into a generative artificial intelligence model.

[0361] If the model detects an anomaly, it generates an anomaly alert.

[0362] Step 6:

[0363] The server notifies the user of the detected anomaly, and the user confirms the anomaly. The push notification is sent using Firebase Cloud Messaging. The input includes the transaction data and sentiment data for which the anomaly was detected. The output is the notification and confirmation for the user.

[0364] Specific behavior:

[0365] Use Firebase Cloud Messaging to send push notifications to users' devices.

[0366] Users can review notifications and approve or report unusual transactions within the app.

[0367] Step 7:

[0368] The server periodically updates the transaction data and shipping information and retrains the generative AI model to improve its accuracy. The input includes new transaction data and updated emotion data. The output is a retrained, highly accurate generative AI model.

[0369] Specific behavior:

[0370] Extract new data from a MySQL database.

[0371] Retrain the model and reevaluate its performance.

[0372] Adjust parameters as needed to improve the accuracy of the model.

[0373] (Application example 2)

[0374] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0375] Fraudulent credit card use is becoming more sophisticated every year, making it difficult to quickly and accurately detect fraudulent transactions using conventional methods. Furthermore, when a fraudulent transaction is detected, notifications are given to users uniformly, and no response is given based on the user's emotional state. This often leaves users feeling anxious and stressed. The present invention aims to solve these problems and realize effective fraud detection and smooth user response.

[0376] The identification processing by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting credit card transaction data and delivery destination information, means for normalizing the transaction data and delivery destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for collecting and analyzing user emotion data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, and means for notifying the user of detected anomalies, customizing the content of the notification based on the user's emotional state, and allowing the user to confirm the anomaly. This enables highly accurate real-time detection of fraudulent transactions and enables appropriate notification and response tailored to the user's emotional state.

[0377] Output for definition statements

[0378] "Credit card transaction data" refers to information about transactions conducted using a credit card, including details such as the transaction date and time, transaction amount, and transaction store.

[0379] "Shipping destination information" refers to information about the shipping destination of the purchased product, specifically information such as address, name, and contact details.

[0380] "Normalization" refers to the process of maintaining data consistency and ensuring consistency across different data sets.

[0381] An "outlier" is a data point that falls outside the normal range, such as an extremely high or low value in a data set.

[0382] "Features" refer to specific attributes or parameters of input data that a generative artificial intelligence model uses for learning and prediction.

[0383] A "generative artificial intelligence model" is a statistical or machine learning model that learns patterns and features based on large amounts of data and makes predictions and classifications.

[0384] "Emotional data" refers to information about the user's emotional state obtained from their voice, facial expression, text input, etc.

[0385] "Real-time" refers to immediate processing or reaction, meaning data is collected, analyzed, and responded to with short latency.

[0386] "Anomaly detection" refers to the process of automatically identifying fraudulent activity or anomalous behavior that deviates from normal usage patterns.

[0387] "Customizing notification content" refers to changing the message and method of notification to a user to suit the status and characteristics of each individual user.

[0388] MODE FOR CARRYING OUT THE INVENTION

[0389] Overall system overview

[0390] The system of the present invention includes components for detecting fraudulent credit card use with high accuracy and for providing prompt and appropriate responses to users. The system is primarily composed of a server, terminals, and users.

[0391] Server Processing

[0392] The server processes the data in the following steps to detect and notify anomalies.

[0393] 1. Data Collection

[0394] The server collects credit card transaction data and shipping information, including details such as transaction date and time, amount, currency, merchant, and shipping address.

[0395] Hardware: The servers are general-purpose servers equipped with high-performance processors and storage.

[0396] Software: Data is collected through a database management system (DBMS) or API.

[0397] 2. Data Preprocessing

[0398] The collected data is normalized, outliers are removed, and features are extracted, providing data suitable for training generative AI models.

[0399] Software used: Data preprocessing scripts using programming languages ​​such as Python and R, and data manipulation libraries such as Pandas and NumPy.

[0400] 3. Learning generative AI models

[0401] The normalized and feature-extracted data is used to train a generative AI model, which learns the characteristics of normal and fraudulent usage patterns.

[0402] Hardware used: High-performance server equipped with GPU etc.

[0403] Software used: Machine learning libraries such as TensorFlow and PyTorch.

[0404] 4. Emotional Data Collection and Analysis

[0405] It collects and analyzes emotional data such as voice and facial expressions from users' smartphones and other devices, uses an emotion engine to identify emotional states, and then applies that data to generative AI models.

[0406] Hardware used: Smartphone camera and microphone.

[0407] Software used: Speech recognition API and image analysis API.

[0408] 5. Real-time detection and notification

[0409] It receives real-time data as transactions occur, analyzes it with a generative AI model, and sends customized notifications based on the user's emotional state if an anomaly is detected.

[0410] Hardware used: High-performance server capable of real-time processing.

[0411] Software used: Real-time data streaming framework, push notification service.

[0412] Specific examples

[0413] When a user makes a transaction using a credit card, the data (e.g., the transaction amount is 150,000 yen, and the transaction location is New York) is sent to the server. The server immediately analyzes this transaction data using a generative AI model and compares it with normal usage patterns. If the transaction is determined to be abnormal, the server uses an emotion engine to obtain emotional data from the smartphone's camera and microphone and detects that the user is anxious. As a result, it can send a notification adapted to the user's emotional state, such as displaying a message saying, "There is a suspicion of a fraudulent transaction. Don't worry, we will take immediate action."

[0414] Prompt Sentence Examples

[0415] Here are some examples of prompts for generative AI models:

[0416] Please determine whether the following credit card transaction is legitimate. Transaction information: Amount: 150,000 yen, Location: New York, User: Tokyo. Is this transaction different from the usual pattern? Also, the user's voice indicates anxiety.

[0417] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0418] Program processing steps

[0419] Step 1:

[0420] The server collects all credit card transaction data and shipping information. For each transaction, details such as transaction date and time, transaction amount, currency, transaction store, shipping address, etc. are stored in a database. This input data is recorded as a record for each transaction.

[0421] Step 2:

[0422] The server normalizes the collected transaction data and removes outliers. This includes maintaining data consistency and standardizing the format. For example, it detects and removes transactions with unusually high prices or inaccurate shipping information. This process produces clean, consistent feature data.

[0423] Step 3:

[0424] Using the normalized and feature-extracted data, the server trains a generative AI model. This model learns past patterns of normal and fraudulent usage. The data inputs are historical transaction data and its features, and the output is feedback that the model uses to classify normal and abnormal behavior.

[0425] Step 4:

[0426] A user's smartphone or other device collects and sends emotional data to a server. Emotional data includes information such as the user's voice and facial expressions. The emotion engine analyzes this data and identifies the user's emotional state (e.g., anxiety, stress). This data is the input data obtained through the emotion engine API, and the analyzed emotional state is the output.

[0427] Step 5:

[0428] The server receives transaction data in real time and analyzes it using a generative AI model. Here, detailed transaction information is the input data, and the output is an anomaly detected as the analysis result of the AI ​​model. If the AI ​​detects an anomaly, the details are immediately passed on to the next step.

[0429] Step 6:

[0430] The server notifies the user of detected anomalies and customizes the notification content based on the user's emotional state. The notification message is generated based on the user's emotional state obtained by the emotion engine and the transaction data. For example, if the server detects that the user is feeling anxious or stressed, the notification content is customized with gentler language such as "Don't worry, we will take immediate action." The output data of the emotion engine and the abnormal transaction data are used as input, and the output is a notification message to the user.

[0431] Step 7:

[0432] The user receives a notification through their device and checks whether the abnormal transaction is legitimate. If the user checks the transaction and determines it to be fraudulent, they can send instructions to the server via the app, such as suspending their credit card or changing the shipping address. This inputs the user's confirmation data, and the terminal then sends instructions to the card company, etc., which is the output.

[0433] Step 8:

[0434] The server periodically updates transaction data and delivery destination information, retraining the generative AI model to maintain or improve its accuracy. Emotional data is also periodically added and incorporated into the model. This allows the system to always use the latest information for highly accurate fraud detection. The latest transaction data and emotional data are used as input, and the output is an updated AI model.

[0435] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0436] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0437] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0438] [Second embodiment]

[0439] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0440] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0441] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0442] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0443] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0444] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0445] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0446] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0447] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0448] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0449] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0450] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0451] This invention is a system that collects credit card transaction data and shipping destination information, and uses a generative artificial intelligence model to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[0452] Data collection and preprocessing

[0453] The server collects credit card transaction data and shipping information for each transaction, as well as known fraud data. The collected data is normalized and outliers are removed. This extracts features and prepares the data for processing.

[0454] Examples:

[0455] When a user purchases a product from an online shop, transaction data (purchase date and time, amount, shop location, etc.) is collected. At the same time, delivery address information for the product is also obtained.

[0456] Model training and generation

[0457] The server uses the preprocessed data to train a generative artificial intelligence model, which learns normal usage patterns and the characteristics of fraudulent usage, and has the ability to detect anomalies with high accuracy.

[0458] Examples:

[0459] The server analyzes past transaction data to learn, for example, which region a user typically spends money in. This allows the model to memorize typical usage patterns.

[0460] Real-time detection

[0461] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, an alert is generated immediately.

[0462] Examples:

[0463] If a user makes a sudden, expensive purchase in a particular major city while traveling abroad, the transaction data is sent to the server, and the model determines that it differs from the normal pattern, detecting the anomaly and generating a warning.

[0464] User notification and response

[0465] If the server detects an abnormality, it will immediately notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user will take steps to suspend the card or change the shipping address.

[0466] Examples:

[0467] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and, for example, determines that the transaction was not initiated by them, they can press a button in the app to immediately report the incident to their credit card company and take steps to suspend their card.

[0468] Continuous model updates

[0469] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy, allowing the system to always perform highly accurate fraud detection based on the latest data.

[0470] Examples:

[0471] Through monthly data updates, the server learns new fraud patterns and retrains the model, keeping the system up to date with the latest fraud techniques.

[0472] In this way, the system of the present invention detects fraudulent credit card use and fraudulent changes to shipping addresses in real time with high accuracy, and provides prompt notification and response to users, thereby realizing a safe and secure transaction environment for users.

[0473] The processing flow will be explained below.

[0474] Step 1:

[0475] The server periodically collects credit card transaction data and shipping information.

[0476] Specific operation: Using the APIs of credit card companies and delivery companies, the system obtains the latest transaction data and delivery address information, while also collecting data on known past fraudulent use.

[0477] Step 2:

[0478] The server normalizes the collected data, removes outliers, and extracts features.

[0479] Specific operations: Extracts key fields such as date, time, amount, store used, and delivery destination from raw data and normalizes them into a consistent format. Detects and removes outliers and calculates the features required for analysis.

[0480] Step 3:

[0481] The server trains a generative artificial intelligence model using the preprocessed data.

[0482] Specific operations: Split the normalized and feature-extracted data into a training set and a test set, and train a generative AI model to learn normal usage patterns and the characteristics of fraudulent usage.

[0483] Step 4:

[0484] When a user makes a transaction using a credit card, the terminal transmits the transaction data to the server in real time.

[0485] What it does: The POS system or online payment platform instantly sends the user's transaction data (e.g., time, location, amount, etc.) to the server.

[0486] Step 5:

[0487] The server immediately analyzes the received transaction data using a generative artificial intelligence model and detects any anomalies.

[0488] How it works: Incoming data is fed into a generative AI model in real time, where it is compared with normal usage patterns. If an abnormal pattern is detected, an anomaly flag is raised.

[0489] Step 6:

[0490] If an abnormality is detected, the server will promptly notify the user.

[0491] Specific behavior: If an anomaly is flagged, a warning message will be sent using the user's contact information (email address, SMS, push notification, etc.).

[0492] Step 7:

[0493] The user receives a notification and verifies the authenticity of the transaction.

[0494] Specific operation: When a user receives a notification, they access a web portal or smartphone app to verify whether the transaction was initiated by them.

[0495] Step 8:

[0496] If the user determines that the use has been fraudulent, procedures such as suspending the card or changing the delivery address will be implemented.

[0497] Specific actions: If a user determines that a transaction has been fraudulent, they can contact their credit card company using a link in the dedicated app or web portal to suspend their card. They can also contact the delivery company to suspend or change the delivery of their product.

[0498] Step 9:

[0499] The server periodically updates transaction data and delivery destination information and retrains the generative artificial intelligence model.

[0500] What it does: Periodically prepare new datasets to evaluate the performance of existing generative AI models, retraining them as needed to maintain or improve their accuracy.

[0501] Through the above steps, the system of the present invention detects credit card fraud and shipping address fraud in real time, promptly notifying the user so that appropriate action can be taken.

[0502] Example 1

[0503] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0504] Fraudulent credit card use can cause serious financial damage to users. Early detection of fraudulent use and appropriate countermeasures are required, but current systems have difficulty detecting fraud in real time with high accuracy. Furthermore, maintaining the accuracy of the system requires continuous model updates, which are difficult to achieve efficiently with conventional methods.

[0505] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0506] In this invention, the server includes means for collecting credit card transaction data and recipient information, means for normalizing the transaction data and recipient information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, means for notifying the user of detected anomalies, allowing the user to confirm the anomaly and, if fraudulent, suspending the payment method or changing the recipient information, and means for periodically updating the transaction data and recipient information and relearning the generative artificial intelligence model to maintain or improve its accuracy. This makes it possible to detect fraudulent credit card use in real time with high accuracy and provide users with prompt notification and support.

[0507] A "credit card" is a payment method issued by a bank or credit card company that users use when purchasing goods or services.

[0508] "Transaction data" refers to information relating to purchases or payments made using a credit card, including, for example, the date and time of purchase, the amount, and the place of purchase.

[0509] "Recipient information" refers to information about the delivery address of purchased products or services, and is data including information such as the address, recipient name, and contact details.

[0510] "Normalization" is the process of converting data values ​​to a uniform scale to ensure data consistency and comparability.

[0511] An "outlier" is a value that deviates from the normal data pattern and should be treated as noise or an error in data analysis.

[0512] In data analysis and machine learning, a "feature" is a numerical value or indicator that represents an important attribute or pattern of data.

[0513] A "generative artificial intelligence model" is a type of artificial intelligence that is an algorithm that learns patterns and characteristics of data and makes predictions and detects anomalies in new data.

[0514] "Real-time" refers to processing data either instantly as it occurs or with very little delay.

[0515] "Anomaly detection" is the process of identifying data that deviates from normal patterns, indicating potential problems or fraud.

[0516] "User" means an end user who purchases goods or services using a credit card and receives notifications from the system.

[0517] "Payment Instrument" means a method of paying for goods and services, including credit cards and debit cards.

[0518] "Model retraining" is the process of using new collected data to readjust the model parameters and algorithms in order to maintain or improve the accuracy of a generative artificial intelligence model.

[0519] This invention is a system that collects credit card transaction data and recipient information, and uses a generative artificial intelligence model to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[0520] Data collection and preprocessing

[0521] The server collects credit card transaction data and payee information for each transaction, as well as known fraud data. Through an API, the server normalizes the collected data and removes outliers. This extracts features and prepares the data for processing.

[0522] Examples:

[0523] For example, when a user purchases a product from an online shop, the transaction data (purchase date and time, amount, shop location, etc.) is sent to the server. At the same time, the server also collects information about the recipient of the product.

[0524] Model learning

[0525] The server uses the preprocessed data to train a generative AI model, which then learns normal usage patterns and the characteristics of fraudulent usage, detecting anomalies with high accuracy.

[0526] Examples:

[0527] The server analyzes past transaction data to learn, for example, which region a user typically spends money in. This allows the server to memorize typical usage patterns into a model.

[0528] Real-time detection

[0529] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, an alert is generated immediately.

[0530] Examples:

[0531] For example, if a user makes a large purchase in a particular major city while traveling abroad, the transaction data is sent to the server, and the model determines that the transaction differs from the normal pattern. As a result, an anomaly is detected and an alert is generated.

[0532] User notification and response

[0533] If the server detects an abnormality, it will immediately notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user can take steps to stop the payment method or change the payee information.

[0534] Examples:

[0535] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and, for example, determines that the transaction was not made by them, they can press a button in the app to immediately report it to their credit card company and take steps to suspend their credit card.

[0536] Continuous model updates

[0537] The server periodically updates transaction data and recipient information, and retrains the generative AI model to maintain or improve its accuracy, allowing the system to always perform highly accurate fraud detection based on the latest data.

[0538] Examples:

[0539] Every month, the server retrains the model using a new dataset collected, updating the model parameters to the latest state, thereby keeping the system adaptable to new fraud techniques.

[0540] Prompt Sentence Examples

[0541] "Please retrain your fraud detection model using this month's transaction data."

[0542] This system makes it possible to detect fraudulent credit card use in real time with high accuracy, and to provide users with prompt notification and response.

[0543] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0544] Step 1:

[0545] Data collection

[0546] The server collects transaction data and recipient information from the APIs of credit card companies and online shops. Inputs include the purchase date and time, amount, shop location, and recipient information for each transaction. This data is stored on the server.

[0547] Specific behavior:

[0548] When a new transaction occurs, the server receives the transaction data in real time through the API. For example, when a user purchases a product, the transaction information is sent to the server.

[0549] Step 2:

[0550] Fraud data collection

[0551] The server also collects known fraud data. The input is transaction data of previously detected fraudulent activity. This data serves as the basis for learning fraud patterns.

[0552] Specific behavior:

[0553] The server periodically downloads fraud data from credit card companies and other data providers.

[0554] Step 3:

[0555] Data normalization and outlier removal

[0556] The server normalizes the collected transaction data and recipient information and removes outliers. The input is the collected raw data, and the output is the normalized data. This keeps the data consistent and makes it easier to analyze.

[0557] Specific behavior:

[0558] Check the transaction dataset stored in the database and perform preprocessing such as scaling and missing value imputation.

[0559] Step 4:

[0560] Feature extraction

[0561] The server extracts features from the preprocessed data. The input is normalized data, and the output is a set of features. The features include transaction frequency, amount statistics, geographic information, etc.

[0562] Specific behavior:

[0563] For each transaction, multiple metrics (purchase date and time, average and variance of the amount, geographical distance, etc.) are calculated to create a feature set.

[0564] Step 5:

[0565] Learning generative artificial intelligence models

[0566] The server uses the extracted features to train a generative AI model. The input is a set of features, and the output is a trained generative AI model. This allows the server to learn the characteristics of normal usage patterns and fraudulent usage.

[0567] Specific behavior:

[0568] Historical trading data is used as a training set to optimize the model parameters.

[0569] Step 6:

[0570] Real-time data transmission

[0571] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The input is the user's transaction data, and the transmitted data arrives at the server.

[0572] Specific behavior:

[0573] The user's purchase information is immediately sent from the terminal to the server.

[0574] Step 7:

[0575] Real-time analytics

[0576] The server instantly analyzes the received transaction data using a generative artificial intelligence model. The input is the transaction data sent in real time, and the output is the presence or absence of anomalies. If there is a possibility of fraud, a warning is generated.

[0577] Specific behavior:

[0578] The data is analyzed layer by layer, and when an anomaly is detected, an anomaly score is generated.

[0579] Step 8:

[0580] User Notifications

[0581] If an abnormality is detected, the server will promptly send a notification to the user. The input is the abnormality detection result, and the output is a notification message to the user.

[0582] Specific behavior:

[0583] When an anomaly is detected, the server generates a notification message and sends a push notification to the user's smartphone.

[0584] Step 9:

[0585] User Support

[0586] The user receives a notification and verifies whether the transaction is legitimate. If it is fraudulent, they can block the payment method or change the payee information. The input is the user's confirmation, and the output is an action such as reporting to the card company.

[0587] Specific behavior:

[0588] If the user checks the notification and presses the "This transaction was not made by me" button, the transaction will be reported to the credit card company and the credit card will be suspended.

[0589] Step 10:

[0590] Model Update

[0591] The server periodically updates the transaction data and recipient information and retrains the generative AI model to maintain or improve its accuracy. The input is a new dataset and the output is an updated generative AI model.

[0592] Specific behavior:

[0593] Every month, the model is retrained using the latest collected data and the model parameters are optimized.

[0594] (Application example 1)

[0595] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0596] Conventional credit card fraud detection systems were slow to detect fraud, with users often only realizing they had been compromised after the fraud had actually occurred. Furthermore, they lacked the functionality to allow users to respond quickly after fraud was detected, making it difficult to prevent fraudulent use. Furthermore, as fraud patterns change, continuous system updates are required, which was difficult to achieve with conventional systems.

[0597] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0598] In this invention, the server includes means for collecting credit card transaction data and shipping destination information, means for normalizing the transaction data and shipping destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for receiving the transaction data in real time and detecting anomalies using the generative artificial intelligence model, means for sending a push notification of the detected anomaly to the user's communication terminal, and means for allowing the user to immediately take steps to suspend their credit card based on the results of the generative artificial intelligence model. This enables highly accurate detection of fraudulent use in real time and provides an environment in which users can respond quickly.

[0599] A "credit card" is a payment method with a certain credit limit that is used when purchasing goods and services.

[0600] "Transaction Data" refers to detailed information about purchases made using a credit card, including transaction date and time, amount, store location, and other data.

[0601] "Shipping destination information" refers to information about the shipping destination of products purchased with a credit card, and includes the address and name of the recipient.

[0602] "Normalization" is the process of converting data into a consistent format, a technique that makes data analysis and model training easier by standardizing the scale of variables.

[0603] An "outlier" is a piece of data that is statistically significantly different from other data points, and is a number that may affect the results of an analysis.

[0604] "Features" are input data provided to a machine learning model that represent characteristics or attributes that are important for the model to learn data patterns.

[0605] A "generative artificial intelligence model" is a machine learning algorithm designed to learn the patterns and rules needed to perform a specific task, and is typically powered by large amounts of training data.

[0606] "Learning" is the process by which a generative artificial intelligence model finds patterns and rules in given data and improves its performance.

[0607] "Real-time" refers to a method in which data is processed as it is acquired and results are provided immediately.

[0608] "Anomaly detection" is a function that finds behaviors or values ​​in data that differ from normal patterns, and is used to identify fraudulent use or abnormal behavior.

[0609] A "push notification" is a notification message that an application sends directly to a user's communication terminal, and is a means of delivering information to the user immediately.

[0610] "Credit card suspension procedure" is a process to temporarily or permanently disable the use of a credit card suspected of fraudulent use, and is carried out to prevent further damage.

[0611] The system for realizing this invention collects credit card transaction data and shipping destination information, detects fraudulent use in real time using a generative artificial intelligence model, and notifies users. This system is composed of a server, terminals, and users.

[0612] Data collection and preprocessing

[0613] The server collects credit card transaction data and shipping address information for each transaction. The collected data is normalized and outliers are removed. This allows features to be extracted and the data is ready for processing. For example, when a user purchases a product online, the transaction data (purchase date and time, amount, purchase location, etc.) is collected. At the same time, the product's shipping address information is also obtained.

[0614] Model learning

[0615] The server uses the preprocessed data to train a generative artificial intelligence model. This model learns the characteristics of normal usage patterns and fraudulent usage, and has the ability to detect anomalies with high accuracy. For example, the server analyzes past transaction data to learn which region a user is shopping in and within what price range. This allows normal usage patterns to be memorized in the model.

[0616] Real-time detection

[0617] When a user purchases a product, the user's device sends the transaction data to a server in real time. The data received by the server is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, a warning is immediately generated. For example, if a user makes an expensive purchase in an unusual location while traveling abroad, the transaction data is sent to the server and is determined to be abnormal by the model. As a result, the anomaly is immediately detected and a warning is generated.

[0618] User notification and response

[0619] If an abnormality is detected, the server will promptly send a notification to the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user can take steps to suspend their credit card. For example, when the server detects an abnormality, a push notification will be sent to the user's smartphone. If the user checks the notification and determines that the transaction was not made by them, they can press a button in the app to report it to their credit card company and take steps to suspend their card.

[0620] Continuous model updates

[0621] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy. This allows the system to always perform highly accurate fraud detection based on the latest data. For example, the server updates data monthly, learns new fraudulent usage patterns, and retrains the model. This makes it possible to respond to the latest fraud techniques.

[0622] Prompt Sentence Examples

[0623] Examples of prompts that can be provided to generative AI models include:

[0624] "Does the purchase of over 5,000 yen in Shinjuku Ward match a pattern that has not been identified as abnormal in your past transaction history?"

[0625] The above is a specific embodiment of the present invention. This system makes it possible to detect fraudulent use with high accuracy in real time and to provide an environment in which users can respond quickly.

[0626] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0627] Step 1:

[0628] The server collects credit card transaction data and shipping information. Specifically, it obtains detailed information about each transaction (transaction date and time, amount, store location, shipping information, etc.) through a data collection API. The input is the detailed information about each transaction, and the output is a list of the collected transaction data and shipping information.

[0629] Step 2:

[0630] The server normalizes the collected transaction data and shipping destination information and removes outliers. Specifically, it performs a data cleansing process to remove outliers and missing values ​​and format the data into a unified format. The input is a list of collected transaction data, and the output is the normalized and cleansed data.

[0631] Step 3:

[0632] The server extracts features from the normalized and cleansed data. Specifically, it calculates and extracts features for each transaction by amount, date, time, and region. The input is the normalized and cleansed data, and the output is the data with extracted features.

[0633] Step 4:

[0634] The server trains a generative AI model using the data from which features have been extracted. Specifically, it performs model training to learn normal transaction patterns and fraudulent usage patterns based on past transaction data. The input is the data from which features have been extracted, and the output is the trained generative AI model.

[0635] Step 5:

[0636] When a new transaction occurs, the user's device sends the transaction data to the server in real time. Specifically, the data transmission API is called the moment the transaction is completed. The input is the new transaction data, and the output is the transaction data sent to the server.

[0637] Step 6:

[0638] The server analyzes the transaction data received in real time using a generative artificial intelligence model to detect anomalies. Specifically, the transaction data is input into the model, and the results are judged to be abnormal. The input is the transaction data received in real time and the trained model, and the output is the judgment result of whether or not there is an abnormality.

[0639] Step 7:

[0640] If an anomaly is detected, the server promptly sends a push notification to the user's device. Specifically, it uses the notification service API to send the details of the anomaly detection to the user's device. The input is the detected anomaly information, and the output is the push notification sent to the user.

[0641] Step 8:

[0642] The user receives a push notification on their device and verifies the legitimacy of the transaction. If they determine that the transaction is fraudulent, they can proceed with suspending their credit card through the application. Specifically, by pressing a button within the application, fraudulent use is reported and the card is suspended. The input is the user's confirmation result, and the output is the completion of the card suspension procedure.

[0643] Step 9:

[0644] The server periodically updates transaction data and delivery destination information, and retrains the generative AI model to improve its accuracy. Specifically, it periodically updates data and trains it to learn new fraudulent usage patterns. The input is the latest transaction data, and the output is the updated generative AI model.

[0645] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0646] This invention is a system that collects credit card transaction data and shipping destination information, and uses a generative AI model combined with an emotion engine to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[0647] Data collection and preprocessing

[0648] The server collects credit card transaction data and shipping information for each transaction, as well as known fraud data. The collected data is normalized and outliers are removed. This extracts features and prepares the data for processing.

[0649] Examples:

[0650] When a user purchases a product from an online shop, transaction data (purchase date and time, amount, shop location, etc.) is collected. At the same time, delivery address information for the product is also obtained.

[0651] Model training and generation

[0652] The server uses the preprocessed data to train a generative AI model. This model learns normal usage patterns and the characteristics of fraudulent usage, and is capable of detecting anomalies with high accuracy. Meanwhile, the emotion engine recognizes emotions from the user's voice, facial expressions, and text input, and analyzes the data. Emotional data is also used to train the generative AI model.

[0653] Examples:

[0654] The server analyzes past transaction data to learn, for example, which region a user typically shopped in and within what price range. This allows the model to memorize typical usage patterns. At the same time, it also analyzes user emotional data and incorporates it into the model.

[0655] Real-time detection

[0656] When a user purchases a product, the device sends the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an abnormality is detected, an alert is generated immediately. The user's reaction is also analyzed by an emotion engine, and the user's level of stress and anxiety is also detected.

[0657] Examples:

[0658] If a user makes a sudden, expensive purchase in a specific major city while traveling abroad, the transaction data is sent to the server, and the model determines that it differs from normal patterns. As a result, an anomaly is detected, and analysis by the emotion engine detects the user's feelings of surprise or anxiety.

[0659] User notification and response

[0660] If an abnormality is detected, the server will promptly notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user will go through the necessary procedures to suspend the card or change the shipping address. The emotion engine analyzes the user's reaction and customizes the notification content and response procedures as needed.

[0661] Examples:

[0662] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and identifies that the transaction was not initiated by them, the emotion engine detects that the user is extremely anxious and provides gentle instructions on how to respond. The user can also press a button in the app to immediately report the incident to their credit card company and take steps to suspend their card.

[0663] Continuous model updates

[0664] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy. This allows the system to always perform highly accurate fraud detection based on the latest data. Furthermore, emotion data from an emotion engine is added, enabling more personalized responses.

[0665] Examples:

[0666] The server learns new fraud patterns through monthly data updates and retrains the model, ensuring the system can respond to the latest fraud techniques. Data from the emotion engine is also incorporated into the model, making it possible to customize notification methods and countermeasures based on the user's emotional state.

[0667] In this way, the system of the present invention detects fraudulent credit card use and fraudulent changes to shipping addresses in real time with high accuracy, and provides prompt notification and response that takes into account the user's emotional state, thereby realizing a safe and secure transaction environment for users and reducing their mental burden.

[0668] The processing flow will be explained below.

[0669] Step 1:

[0670] The server periodically collects credit card transaction data and shipping information.

[0671] Specific operation: Using the APIs of credit card companies and delivery companies, the system obtains the latest transaction data and delivery address information, while also collecting data on known past fraudulent use.

[0672] Step 2:

[0673] The server normalizes the collected data, removes outliers, and extracts features.

[0674] Specific operations: Extract key fields such as date, time, amount, store used, and delivery destination from raw data, normalize them into a consistent format, detect and remove outliers, and calculate the features required for analysis.

[0675] Step 3:

[0676] The server trains a generative artificial intelligence model using the preprocessed data.

[0677] Specific operation: The normalized and feature-extracted data is split into a training set and a test set. A generative AI model is trained to learn normal usage patterns and the characteristics of fraudulent usage.

[0678] Step 4:

[0679] When a user makes a transaction using a credit card, the terminal transmits the transaction data to the server in real time.

[0680] Specific operation: The POS system or online payment platform immediately sends the user's transaction data (e.g., time, location, amount, etc.) to the server.

[0681] Step 5:

[0682] The server immediately analyzes the received transaction data using a generative artificial intelligence model and detects any anomalies.

[0683] How it works: Incoming data is fed into a generative AI model in real time, where it is compared with normal usage patterns. If an abnormal pattern is detected, an anomaly flag is raised.

[0684] Step 6:

[0685] If an abnormality is detected, the server will promptly notify the user.

[0686] Specific behavior: If an abnormality flag is raised, a warning message will be sent using the user's contact information (email address, SMS, push notification, etc.).

[0687] Step 7:

[0688] The emotion engine analyzes the user's reaction in real time when they receive a notification and detects their emotions (e.g., surprise, anxiety, anger, etc.).

[0689] Specific operation: The emotion engine analyzes the user's voice, facial expressions, text input, etc. when receiving the notification to determine their emotional state.

[0690] Step 8:

[0691] The user receives a notification and verifies the authenticity of the transaction.

[0692] Specific operation: Upon receiving the notification, the user accesses the web portal or smartphone app to verify whether the transaction was initiated by them.

[0693] Step 9:

[0694] The server customizes the notification content and response procedures based on the user's emotions detected by the emotion engine.

[0695] Specific behavior: If the user's emotions indicate anxiety or surprise, the server will re-notify them with softer language and reassuring messages, and if necessary, instruct them to escalate the issue to a support staff member.

[0696] Step 10:

[0697] If the user determines that the use has been fraudulent, procedures such as suspending the card or changing the delivery address will be implemented.

[0698] Specific operation: If a user determines that a transaction has been fraudulent, they can contact their credit card company using a dedicated app or a link in the web portal to suspend their card. They can also contact the delivery company to suspend or change the delivery of their product.

[0699] Step 11:

[0700] The server periodically updates transaction data and delivery destination information and retrains the generative artificial intelligence model.

[0701] How it works: Regularly prepare new datasets and evaluate the performance of existing generative AI models. Retrain as needed to maintain or improve model accuracy. Additionally, by adding data from the emotion engine, the model can be trained to provide more personalized responses.

[0702] Through these steps, the system of the present invention detects credit card fraud and shipping fraud in real time and enables quick and appropriate responses based on the user's emotional state.

[0703] Example 2

[0704] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0705] In modern society, credit card fraud is on the rise, with the risk increasing especially in online shopping and remote transactions. When fraud occurs, it not only causes financial loss to users, but also causes significant psychological stress. Furthermore, existing fraud detection systems are often inadequate because they struggle to respond in real time and are unable to take into account the user's emotional state. This makes it difficult for many users to conduct transactions with confidence. Therefore, there is a need for a system that can detect fraud quickly and accurately, and provide notifications and responses that take the user's emotions into account.

[0706] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for collecting credit card transaction data and delivery destination information, means for normalizing the transaction data and delivery destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for collecting and analyzing user emotion data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, and means for notifying the user of detected anomalies and allowing the user to confirm the anomaly. This makes it possible to quickly and accurately detect fraudulent credit card use and to notify and respond in consideration of the user's emotional state.

[0707] "Credit card transaction data" means information that records purchases and payments made using a credit card.

[0708] "Delivery destination information" refers to information such as the address and contact details of the recipient of the purchased product or service.

[0709] "Normalization" is the process of standardizing the range and units of data to make it easier to compare and analyze.

[0710] "Outlier removal" is the process of identifying and eliminating invalid data points within a data set.

[0711] "Features" are important input variables used in machine learning models for prediction and classification.

[0712] A "generative artificial intelligence model" is an AI model that has the ability to learn from data, recognize patterns, and generate or analyze new data.

[0713] "User emotion data" refers to information about the user's emotional state obtained from the user's facial expression, voice, text input, and the like.

[0714] "Receiving transaction data in real time" means that the data is sent to the server and received immediately the moment a transaction occurs.

[0715] "Anomaly detection" refers to identifying fraudulent or unusual behavior that deviates from normal patterns.

[0716] "Notification" is the act of the system informing the user of abnormalities or important information.

[0717] This invention relates to a highly accurate and rapid detection system for preventing fraudulent use of credit cards. This system collects credit card transaction data and shipping destination information, and uses this data to learn and apply a generative artificial intelligence model that detects fraudulent use. The details are as follows:

[0718] Data collection and preprocessing

[0719] The server collects credit card transaction data and shipping information for each transaction, which involves retrieving the data through an API and storing it in a database using MySQL, a popular relational database management system (RDBMS).

[0720] Specifically, when a user purchases an item from an online shop, the transaction data (purchase date and time, amount, store location, etc.) and shipping information are stored in a MySQL database. The collected data is read using the Python Pandas library, normalized, and outliers are removed, preparing the data for processing.

[0721] Model training and generation

[0722] The server uses the preprocessed data to train a generative artificial intelligence model. This model learns normal usage patterns and the characteristics of fraudulent usage, and detects anomalies with high accuracy. The specific software used is Scikit-learn's RandomForestClassifier.

[0723] The model training uses transaction data and fraudulent use data collected to date. Additionally, Affectiva's SDK is used as a sentiment analysis library to collect and analyze user sentiment data. This sentiment data is also fed back into the model training.

[0724] Specifically, the server analyzes past transaction data to learn, for example, which region a user typically spends money in. At the same time, it analyzes emotional data using Affectiva's SDK and incorporates that data into the model.

[0725] Real-time detection

[0726] When a user purchases a product, the device sends the transaction data to the server in real time. The server immediately analyzes the received data using a generative artificial intelligence model and immediately generates a warning if an abnormality is detected. The server also analyzes the user's reactions using an emotion engine to detect the user's level of stress and anxiety.

[0727] Specifically, if a user makes a sudden, expensive purchase in a specific major city while traveling abroad, the transaction data is sent to the server, and the model determines that the transaction is different from the usual pattern. If it determines that fraudulent use is suspected, emotion analysis can also detect the user's anxiety.

[0728] User notification and response

[0729] The server uses Firebase Cloud Messaging as a communication library to promptly notify the user if an abnormality is detected. The user who receives the notification can check whether the transaction is legitimate, and if it is fraudulent, can immediately take action such as suspending the card or changing the delivery address.

[0730] Specifically, when the server detects an abnormality, it sends a push notification to the user's smartphone. If the user checks the notification and determines that the transaction was not initiated by them, the app uses emotional analysis to detect that the user is extremely anxious and provides gentle guidance. The app also allows the user to press a button to report the transaction to their credit card company and immediately take steps to suspend their card.

[0731] Continuous model updates

[0732] The server periodically updates transaction data and delivery destination information, and performs re-training to improve the accuracy of the generative AI model. This ensures highly accurate fraud detection based on the latest data. Emotional data is also added, enabling more personalized responses.

[0733] Specifically, the server updates data monthly, learns new fraudulent usage patterns, and retrains the model. This ensures the system can respond to the latest fraud techniques. Emotional data is also fed back to the model, allowing it to provide notification methods and countermeasures tailored to the user.

[0734] Prompt Sentence Examples

[0735] Fraud detection is performed on newly collected transaction data, including:

[0736] Purchase date: 2023-10-01 10:00:00

[0737] Amount: $500

[0738] Shop location: London

[0739] Delivery Address: 123 ABC Street, London

[0740] User emotion data: Moderate anxiety

[0741] Use this data to detect anomalies in real time using fraud detection models and notify users as needed.

[0742] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0743] System program processing flow

[0744] Step 1:

[0745] The server collects credit card transaction data and shipping information. Specifically, it uses an API to retrieve credit card transaction information and stores it in a MySQL database. The input includes the transaction data and shipping information retrieved from the API. The output is this raw data stored in the database.

[0746] Specific behavior:

[0747] Make API calls to retrieve transaction data and shipping information.

[0748] The retrieved data is stored in a MySQL database.

[0749] Step 2:

[0750] The server pre-processes the collected data. It uses the Pandas library to normalize the data, remove outliers, and extract features. The input includes raw data stored in a MySQL database. The output is the pre-processed, feature-extracted data.

[0751] Specific behavior:

[0752] Use Pandas to read data from a MySQL database.

[0753] Normalize the data to make it consistent across ranges.

[0754] Detect and remove outliers.

[0755] Important features (e.g., purchase date and time, amount, store location) are extracted and saved as structured data.

[0756] Step 3:

[0757] The server uses the preprocessed data to train a generative AI model. It uses Scikit-learn's RandomForestClassifier to learn normal transaction patterns and fraudulent usage patterns. The input includes training data with extracted features. The output is a trained generative AI model.

[0758] Specific behavior:

[0759] Define a model using Scikit-learn's RandomForestClassifier.

[0760] The data from which the features have been extracted is supplied to the model as training data and fitted.

[0761] The model is evaluated and parameters are adjusted as necessary.

[0762] Step 4:

[0763] The server collects and analyzes the user's emotional data. Using Affectiva's SDK, it analyzes the user's emotions (e.g., stress, anxiety, surprise, etc.) and stores the data. Inputs include the user's facial expressions, voice, and text input. The output is the analyzed emotional data.

[0764] Specific behavior:

[0765] Affectiva's SDK is used to analyze the user's facial expressions, voice, and text input.

[0766] The analysis results (emotion data) are saved as structured data and provided to a generative artificial intelligence model.

[0767] Step 5:

[0768] When a user purchases a product, the device sends transaction data to a server in real time. The server analyzes the received data using a generative artificial intelligence model to detect anomalies. The input includes real-time transaction data and a pre-trained model. The output is a warning message or notification when an anomaly is detected.

[0769] Specific behavior:

[0770] The terminal collects transaction data and immediately transmits it to the server.

[0771] The server receives the transaction data and inputs it into a generative artificial intelligence model.

[0772] If the model detects an anomaly, it generates an anomaly alert.

[0773] Step 6:

[0774] The server notifies the user of the detected anomaly, and the user confirms the anomaly. The push notification is sent using Firebase Cloud Messaging. The input includes the transaction data and sentiment data for which the anomaly was detected. The output is the notification and confirmation for the user.

[0775] Specific behavior:

[0776] Use Firebase Cloud Messaging to send push notifications to users' devices.

[0777] Users can review notifications and approve or report unusual transactions within the app.

[0778] Step 7:

[0779] The server periodically updates the transaction data and shipping information and retrains the generative AI model to improve its accuracy. The input includes new transaction data and updated emotion data. The output is a retrained, highly accurate generative AI model.

[0780] Specific behavior:

[0781] Extract new data from a MySQL database.

[0782] Retrain the model and reevaluate its performance.

[0783] Adjust parameters as needed to improve the accuracy of the model.

[0784] (Application example 2)

[0785] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0786] Fraudulent credit card use is becoming more sophisticated every year, making it difficult to quickly and accurately detect fraudulent transactions using conventional methods. Furthermore, when a fraudulent transaction is detected, notifications are given to users uniformly, and no response is given based on the user's emotional state. This often leaves users feeling anxious and stressed. The present invention aims to solve these problems and realize effective fraud detection and smooth user response.

[0787] The identification processing by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting credit card transaction data and delivery destination information, means for normalizing the transaction data and delivery destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for collecting and analyzing user emotion data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, and means for notifying the user of detected anomalies, customizing the content of the notification based on the user's emotional state, and allowing the user to confirm the anomaly. This enables highly accurate real-time detection of fraudulent transactions and enables appropriate notification and response tailored to the user's emotional state.

[0788] Output for definition statements

[0789] "Credit card transaction data" refers to information about transactions conducted using a credit card, including details such as the transaction date and time, transaction amount, and transaction store.

[0790] "Shipping destination information" refers to information about the shipping destination of the purchased product, specifically information such as address, name, and contact details.

[0791] "Normalization" refers to the process of maintaining data consistency and ensuring consistency across different data sets.

[0792] An "outlier" is a data point that falls outside the normal range, such as an extremely high or low value in a data set.

[0793] "Features" refer to specific attributes or parameters of input data that a generative artificial intelligence model uses for learning and prediction.

[0794] A "generative artificial intelligence model" is a statistical or machine learning model that learns patterns and features based on large amounts of data and makes predictions and classifications.

[0795] "Emotional data" refers to information about the user's emotional state obtained from their voice, facial expression, text input, etc.

[0796] "Real-time" refers to immediate processing or reaction, meaning data is collected, analyzed, and responded to with short latency.

[0797] "Anomaly detection" refers to the process of automatically identifying fraudulent activity or anomalous behavior that deviates from normal usage patterns.

[0798] "Customizing notification content" refers to changing the message and method of notification to a user to suit the status and characteristics of each individual user.

[0799] MODE FOR CARRYING OUT THE INVENTION

[0800] Overall system overview

[0801] The system of the present invention includes components for detecting fraudulent credit card use with high accuracy and for providing prompt and appropriate responses to users. The system is primarily composed of a server, terminals, and users.

[0802] Server Processing

[0803] The server processes the data in the following steps to detect and notify anomalies.

[0804] 1. Data Collection

[0805] The server collects credit card transaction data and shipping information, including details such as transaction date and time, amount, currency, merchant, and shipping address.

[0806] Hardware: The servers are general-purpose servers equipped with high-performance processors and storage.

[0807] Software: Data is collected through a database management system (DBMS) or API.

[0808] 2. Data Preprocessing

[0809] The collected data is normalized, outliers are removed, and features are extracted, providing data suitable for training generative AI models.

[0810] Software used: Data preprocessing scripts using programming languages ​​such as Python and R, and data manipulation libraries such as Pandas and NumPy.

[0811] 3. Learning generative AI models

[0812] The normalized and feature-extracted data is used to train a generative AI model, which learns the characteristics of normal and fraudulent usage patterns.

[0813] Hardware used: High-performance server equipped with GPU etc.

[0814] Software used: Machine learning libraries such as TensorFlow and PyTorch.

[0815] 4. Emotional Data Collection and Analysis

[0816] It collects and analyzes emotional data such as voice and facial expressions from users' smartphones and other devices, uses an emotion engine to identify emotional states, and then applies that data to generative AI models.

[0817] Hardware used: Smartphone camera and microphone.

[0818] Software used: Speech recognition API and image analysis API.

[0819] 5. Real-time detection and notification

[0820] It receives real-time data as transactions occur, analyzes it with a generative AI model, and sends customized notifications based on the user's emotional state if an anomaly is detected.

[0821] Hardware used: High-performance server capable of real-time processing.

[0822] Software used: Real-time data streaming framework, push notification service.

[0823] Specific examples

[0824] When a user makes a transaction using a credit card, the data (e.g., the transaction amount is 150,000 yen, and the transaction location is New York) is sent to the server. The server immediately analyzes this transaction data using a generative AI model and compares it with normal usage patterns. If the transaction is determined to be abnormal, the server uses an emotion engine to obtain emotional data from the smartphone's camera and microphone and detects that the user is anxious. As a result, it can send a notification adapted to the user's emotional state, such as displaying a message saying, "There is a suspicion of a fraudulent transaction. Don't worry, we will take immediate action."

[0825] Prompt Sentence Examples

[0826] Here are some examples of prompts for generative AI models:

[0827] Please determine whether the following credit card transaction is legitimate. Transaction information: Amount: 150,000 yen, Location: New York, User: Tokyo. Is this transaction different from the usual pattern? Also, the user's voice indicates anxiety.

[0828] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0829] Program processing steps

[0830] Step 1:

[0831] The server collects all credit card transaction data and shipping information. For each transaction, details such as transaction date and time, transaction amount, currency, transaction store, shipping address, etc. are stored in a database. This input data is recorded as a record for each transaction.

[0832] Step 2:

[0833] The server normalizes the collected transaction data and removes outliers. This includes maintaining data consistency and standardizing the format. For example, it detects and removes transactions with unusually high prices or inaccurate shipping information. This process produces clean, consistent feature data.

[0834] Step 3:

[0835] Using the normalized and feature-extracted data, the server trains a generative AI model. This model learns past patterns of normal and fraudulent usage. The data inputs are historical transaction data and its features, and the output is feedback that the model uses to classify normal and abnormal behavior.

[0836] Step 4:

[0837] A user's smartphone or other device collects and sends emotional data to a server. Emotional data includes information such as the user's voice and facial expressions. The emotion engine analyzes this data and identifies the user's emotional state (e.g., anxiety, stress). This data is the input data obtained through the emotion engine API, and the analyzed emotional state is the output.

[0838] Step 5:

[0839] The server receives transaction data in real time and analyzes it using a generative AI model. Here, detailed transaction information is the input data, and the output is an anomaly detected as the analysis result of the AI ​​model. If the AI ​​detects an anomaly, the details are immediately passed on to the next step.

[0840] Step 6:

[0841] The server notifies the user of detected anomalies and customizes the notification content based on the user's emotional state. The notification message is generated based on the user's emotional state obtained by the emotion engine and the transaction data. For example, if the server detects that the user is feeling anxious or stressed, the notification content is customized with gentler language such as "Don't worry, we will take immediate action." The output data of the emotion engine and the abnormal transaction data are used as input, and the output is a notification message to the user.

[0842] Step 7:

[0843] The user receives a notification through their device and checks whether the abnormal transaction is legitimate. If the user checks the transaction and determines it to be fraudulent, they can send instructions to the server via the app, such as suspending their credit card or changing the shipping address. This inputs the user's confirmation data, and the terminal then sends instructions to the card company, etc., which is the output.

[0844] Step 8:

[0845] The server periodically updates transaction data and delivery destination information, retraining the generative AI model to maintain or improve its accuracy. Emotional data is also periodically added and incorporated into the model. This allows the system to always use the latest information for highly accurate fraud detection. The latest transaction data and emotional data are used as input, and the output is an updated AI model.

[0846] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0847] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0848] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0849] [Third embodiment]

[0850] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0851] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0852] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0853] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0854] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0855] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0856] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0857] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0858] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0859] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0860] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0861] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0862] This invention is a system that collects credit card transaction data and shipping destination information, and uses a generative artificial intelligence model to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[0863] Data collection and preprocessing

[0864] The server collects credit card transaction data and shipping information for each transaction, as well as known fraud data. The collected data is normalized and outliers are removed. This extracts features and prepares the data for processing.

[0865] Examples:

[0866] When a user purchases a product from an online shop, transaction data (purchase date and time, amount, shop location, etc.) is collected. At the same time, delivery address information for the product is also obtained.

[0867] Model training and generation

[0868] The server uses the preprocessed data to train a generative artificial intelligence model, which learns normal usage patterns and the characteristics of fraudulent usage, and has the ability to detect anomalies with high accuracy.

[0869] Examples:

[0870] The server analyzes past transaction data to learn, for example, which region a user typically spends money in. This allows the model to memorize typical usage patterns.

[0871] Real-time detection

[0872] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, an alert is generated immediately.

[0873] Examples:

[0874] If a user makes a sudden, expensive purchase in a particular major city while traveling abroad, the transaction data is sent to the server, and the model determines that it differs from the normal pattern, detecting the anomaly and generating a warning.

[0875] User notification and response

[0876] If the server detects an abnormality, it will immediately notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user will take steps to suspend the card or change the shipping address.

[0877] Examples:

[0878] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and, for example, determines that the transaction was not initiated by them, they can press a button in the app to immediately report the incident to their credit card company and take steps to suspend their card.

[0879] Continuous model updates

[0880] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy, allowing the system to always perform highly accurate fraud detection based on the latest data.

[0881] Examples:

[0882] Through monthly data updates, the server learns new fraud patterns and retrains the model, keeping the system up to date with the latest fraud techniques.

[0883] In this way, the system of the present invention detects fraudulent credit card use and fraudulent changes to shipping addresses in real time with high accuracy, and provides prompt notification and response to users, thereby realizing a safe and secure transaction environment for users.

[0884] The processing flow will be explained below.

[0885] Step 1:

[0886] The server periodically collects credit card transaction data and shipping information.

[0887] Specific operation: Using the APIs of credit card companies and delivery companies, the system obtains the latest transaction data and delivery address information, while also collecting data on known past fraudulent use.

[0888] Step 2:

[0889] The server normalizes the collected data, removes outliers, and extracts features.

[0890] Specific operations: Extracts key fields such as date, time, amount, store used, and delivery destination from raw data and normalizes them into a consistent format. Detects and removes outliers and calculates the features required for analysis.

[0891] Step 3:

[0892] The server trains a generative artificial intelligence model using the preprocessed data.

[0893] Specific operations: Split the normalized and feature-extracted data into a training set and a test set, and train a generative AI model to learn normal usage patterns and the characteristics of fraudulent usage.

[0894] Step 4:

[0895] When a user makes a transaction using a credit card, the terminal transmits the transaction data to the server in real time.

[0896] What it does: The POS system or online payment platform instantly sends the user's transaction data (e.g., time, location, amount, etc.) to the server.

[0897] Step 5:

[0898] The server immediately analyzes the received transaction data using a generative artificial intelligence model and detects any anomalies.

[0899] How it works: Incoming data is fed into a generative AI model in real time, where it is compared with normal usage patterns. If an abnormal pattern is detected, an anomaly flag is raised.

[0900] Step 6:

[0901] If an abnormality is detected, the server will promptly notify the user.

[0902] Specific behavior: If an anomaly is flagged, a warning message will be sent using the user's contact information (email address, SMS, push notification, etc.).

[0903] Step 7:

[0904] The user receives a notification and verifies the authenticity of the transaction.

[0905] Specific operation: When a user receives a notification, they access a web portal or smartphone app to verify whether the transaction was initiated by them.

[0906] Step 8:

[0907] If the user determines that the use has been fraudulent, procedures such as suspending the card or changing the delivery address will be implemented.

[0908] Specific actions: If a user determines that a transaction has been fraudulent, they can contact their credit card company using a link in the dedicated app or web portal to suspend their card. They can also contact the delivery company to suspend or change the delivery of their product.

[0909] Step 9:

[0910] The server periodically updates transaction data and delivery destination information and retrains the generative artificial intelligence model.

[0911] What it does: Periodically prepare new datasets to evaluate the performance of existing generative AI models, retraining them as needed to maintain or improve their accuracy.

[0912] Through the above steps, the system of the present invention detects credit card fraud and shipping address fraud in real time, promptly notifying the user so that appropriate action can be taken.

[0913] Example 1

[0914] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0915] Fraudulent credit card use can cause serious financial damage to users. Early detection of fraudulent use and appropriate countermeasures are required, but current systems have difficulty detecting fraud in real time with high accuracy. Furthermore, maintaining the accuracy of the system requires continuous model updates, which are difficult to achieve efficiently with conventional methods.

[0916] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0917] In this invention, the server includes means for collecting credit card transaction data and recipient information, means for normalizing the transaction data and recipient information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, means for notifying the user of detected anomalies, allowing the user to confirm the anomaly and, if fraudulent, suspending the payment method or changing the recipient information, and means for periodically updating the transaction data and recipient information and relearning the generative artificial intelligence model to maintain or improve its accuracy. This makes it possible to detect fraudulent credit card use in real time with high accuracy and provide users with prompt notification and support.

[0918] A "credit card" is a payment method issued by a bank or credit card company that users use when purchasing goods or services.

[0919] "Transaction data" refers to information relating to purchases or payments made using a credit card, including, for example, the date and time of purchase, the amount, and the place of purchase.

[0920] "Recipient information" refers to information about the delivery address of purchased products or services, and is data including information such as the address, recipient name, and contact details.

[0921] "Normalization" is the process of converting data values ​​to a uniform scale to ensure data consistency and comparability.

[0922] An "outlier" is a value that deviates from the normal data pattern and should be treated as noise or an error in data analysis.

[0923] In data analysis and machine learning, a "feature" is a numerical value or indicator that represents an important attribute or pattern of data.

[0924] A "generative artificial intelligence model" is a type of artificial intelligence that is an algorithm that learns patterns and characteristics of data and makes predictions and detects anomalies in new data.

[0925] "Real-time" refers to processing data either instantly as it occurs or with very little delay.

[0926] "Anomaly detection" is the process of identifying data that deviates from normal patterns, indicating potential problems or fraud.

[0927] "User" means an end user who purchases goods or services using a credit card and receives notifications from the system.

[0928] "Payment Instrument" means a method of paying for goods and services, including credit cards and debit cards.

[0929] "Model retraining" is the process of using new collected data to readjust the model parameters and algorithms in order to maintain or improve the accuracy of a generative artificial intelligence model.

[0930] This invention is a system that collects credit card transaction data and recipient information, and uses a generative artificial intelligence model to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[0931] Data collection and preprocessing

[0932] The server collects credit card transaction data and payee information for each transaction, as well as known fraud data. Through an API, the server normalizes the collected data and removes outliers. This extracts features and prepares the data for processing.

[0933] Examples:

[0934] For example, when a user purchases a product from an online shop, the transaction data (purchase date and time, amount, shop location, etc.) is sent to the server. At the same time, the server also collects information about the recipient of the product.

[0935] Model learning

[0936] The server uses the preprocessed data to train a generative AI model, which then learns normal usage patterns and the characteristics of fraudulent usage, detecting anomalies with high accuracy.

[0937] Examples:

[0938] The server analyzes past transaction data to learn, for example, which region a user typically spends money in. This allows the server to memorize typical usage patterns into a model.

[0939] Real-time detection

[0940] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, an alert is generated immediately.

[0941] Examples:

[0942] For example, if a user makes a large purchase in a particular major city while traveling abroad, the transaction data is sent to the server, and the model determines that the transaction differs from the normal pattern. As a result, an anomaly is detected and an alert is generated.

[0943] User notification and response

[0944] If the server detects an abnormality, it will immediately notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user can take steps to stop the payment method or change the payee information.

[0945] Examples:

[0946] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and, for example, determines that the transaction was not made by them, they can press a button in the app to immediately report it to their credit card company and take steps to suspend their credit card.

[0947] Continuous model updates

[0948] The server periodically updates transaction data and recipient information, and retrains the generative AI model to maintain or improve its accuracy, allowing the system to always perform highly accurate fraud detection based on the latest data.

[0949] Examples:

[0950] Every month, the server retrains the model using a new dataset collected, updating the model parameters to the latest state, thereby keeping the system adaptable to new fraud techniques.

[0951] Prompt Sentence Examples

[0952] "Please retrain your fraud detection model using this month's transaction data."

[0953] This system makes it possible to detect fraudulent credit card use in real time with high accuracy, and to provide users with prompt notification and response.

[0954] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0955] Step 1:

[0956] Data collection

[0957] The server collects transaction data and recipient information from the APIs of credit card companies and online shops. Inputs include the purchase date and time, amount, shop location, and recipient information for each transaction. This data is stored on the server.

[0958] Specific behavior:

[0959] When a new transaction occurs, the server receives the transaction data in real time through the API. For example, when a user purchases a product, the transaction information is sent to the server.

[0960] Step 2:

[0961] Fraud data collection

[0962] The server also collects known fraud data. The input is transaction data of previously detected fraudulent activity. This data serves as the basis for learning fraud patterns.

[0963] Specific behavior:

[0964] The server periodically downloads fraud data from credit card companies and other data providers.

[0965] Step 3:

[0966] Data normalization and outlier removal

[0967] The server normalizes the collected transaction data and recipient information and removes outliers. The input is the collected raw data, and the output is the normalized data. This keeps the data consistent and makes it easier to analyze.

[0968] Specific behavior:

[0969] Check the transaction dataset stored in the database and perform preprocessing such as scaling and missing value imputation.

[0970] Step 4:

[0971] Feature extraction

[0972] The server extracts features from the preprocessed data. The input is normalized data, and the output is a set of features. The features include transaction frequency, amount statistics, geographic information, etc.

[0973] Specific behavior:

[0974] For each transaction, multiple metrics (purchase date and time, average and variance of the amount, geographical distance, etc.) are calculated to create a feature set.

[0975] Step 5:

[0976] Learning generative artificial intelligence models

[0977] The server uses the extracted features to train a generative AI model. The input is a set of features, and the output is a trained generative AI model. This allows the server to learn the characteristics of normal usage patterns and fraudulent usage.

[0978] Specific behavior:

[0979] Historical trading data is used as a training set to optimize the model parameters.

[0980] Step 6:

[0981] Real-time data transmission

[0982] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The input is the user's transaction data, and the transmitted data arrives at the server.

[0983] Specific behavior:

[0984] The user's purchase information is immediately sent from the terminal to the server.

[0985] Step 7:

[0986] Real-time analytics

[0987] The server instantly analyzes the received transaction data using a generative artificial intelligence model. The input is the transaction data sent in real time, and the output is the presence or absence of anomalies. If there is a possibility of fraud, a warning is generated.

[0988] Specific behavior:

[0989] The data is analyzed layer by layer, and when an anomaly is detected, an anomaly score is generated.

[0990] Step 8:

[0991] User Notifications

[0992] If an abnormality is detected, the server will promptly send a notification to the user. The input is the abnormality detection result, and the output is a notification message to the user.

[0993] Specific behavior:

[0994] When an anomaly is detected, the server generates a notification message and sends a push notification to the user's smartphone.

[0995] Step 9:

[0996] User Support

[0997] The user receives a notification and verifies whether the transaction is legitimate. If it is fraudulent, they can block the payment method or change the payee information. The input is the user's confirmation, and the output is an action such as reporting to the card company.

[0998] Specific behavior:

[0999] If the user checks the notification and presses the "This transaction was not made by me" button, the transaction will be reported to the credit card company and the credit card will be suspended.

[1000] Step 10:

[1001] Model Update

[1002] The server periodically updates the transaction data and recipient information and retrains the generative AI model to maintain or improve its accuracy. The input is a new dataset and the output is an updated generative AI model.

[1003] Specific behavior:

[1004] Every month, the model is retrained using the latest collected data and the model parameters are optimized.

[1005] (Application example 1)

[1006] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1007] Conventional credit card fraud detection systems were slow to detect fraud, with users often only realizing they had been compromised after the fraud had actually occurred. Furthermore, they lacked the functionality to allow users to respond quickly after fraud was detected, making it difficult to prevent fraudulent use. Furthermore, as fraud patterns change, continuous system updates are required, which was difficult to achieve with conventional systems.

[1008] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1009] In this invention, the server includes means for collecting credit card transaction data and shipping destination information, means for normalizing the transaction data and shipping destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for receiving the transaction data in real time and detecting anomalies using the generative artificial intelligence model, means for sending a push notification of the detected anomaly to the user's communication terminal, and means for allowing the user to immediately take steps to suspend their credit card based on the results of the generative artificial intelligence model. This enables highly accurate detection of fraudulent use in real time and provides an environment in which users can respond quickly.

[1010] A "credit card" is a payment method with a certain credit limit that is used when purchasing goods and services.

[1011] "Transaction Data" refers to detailed information about purchases made using a credit card, including transaction date and time, amount, store location, and other data.

[1012] "Shipping destination information" refers to information about the shipping destination of products purchased with a credit card, and includes the address and name of the recipient.

[1013] "Normalization" is the process of converting data into a consistent format, a technique that makes data analysis and model training easier by standardizing the scale of variables.

[1014] An "outlier" is a piece of data that is statistically significantly different from other data points, and is a number that may affect the results of an analysis.

[1015] "Features" are input data provided to a machine learning model that represent characteristics or attributes that are important for the model to learn data patterns.

[1016] A "generative artificial intelligence model" is a machine learning algorithm designed to learn the patterns and rules needed to perform a specific task, and is typically powered by large amounts of training data.

[1017] "Learning" is the process by which a generative artificial intelligence model finds patterns and rules in given data and improves its performance.

[1018] "Real-time" refers to a method in which data is processed as it is acquired and results are provided immediately.

[1019] "Anomaly detection" is a function that finds behaviors or values ​​in data that differ from normal patterns, and is used to identify fraudulent use or abnormal behavior.

[1020] A "push notification" is a notification message that an application sends directly to a user's communication terminal, and is a means of delivering information to the user immediately.

[1021] "Credit card suspension procedure" is a process to temporarily or permanently disable the use of a credit card suspected of fraudulent use, and is carried out to prevent further damage.

[1022] The system for realizing this invention collects credit card transaction data and shipping destination information, detects fraudulent use in real time using a generative artificial intelligence model, and notifies users. This system is composed of a server, terminals, and users.

[1023] Data collection and preprocessing

[1024] The server collects credit card transaction data and shipping address information for each transaction. The collected data is normalized and outliers are removed. This allows features to be extracted and the data is ready for processing. For example, when a user purchases a product online, the transaction data (purchase date and time, amount, purchase location, etc.) is collected. At the same time, the product's shipping address information is also obtained.

[1025] Model learning

[1026] The server uses the preprocessed data to train a generative artificial intelligence model. This model learns the characteristics of normal usage patterns and fraudulent usage, and has the ability to detect anomalies with high accuracy. For example, the server analyzes past transaction data to learn which region a user is shopping in and within what price range. This allows normal usage patterns to be memorized in the model.

[1027] Real-time detection

[1028] When a user purchases a product, the user's device sends the transaction data to a server in real time. The data received by the server is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, a warning is immediately generated. For example, if a user makes an expensive purchase in an unusual location while traveling abroad, the transaction data is sent to the server and is determined to be abnormal by the model. As a result, the anomaly is immediately detected and a warning is generated.

[1029] User notification and response

[1030] If an abnormality is detected, the server will promptly send a notification to the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user can take steps to suspend their credit card. For example, when the server detects an abnormality, a push notification will be sent to the user's smartphone. If the user checks the notification and determines that the transaction was not made by them, they can press a button in the app to report it to their credit card company and take steps to suspend their card.

[1031] Continuous model updates

[1032] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy. This allows the system to always perform highly accurate fraud detection based on the latest data. For example, the server updates data monthly, learns new fraudulent usage patterns, and retrains the model. This makes it possible to respond to the latest fraud techniques.

[1033] Prompt Sentence Examples

[1034] Examples of prompts that can be provided to generative AI models include:

[1035] "Does the purchase of over 5,000 yen in Shinjuku Ward match a pattern that has not been identified as abnormal in your past transaction history?"

[1036] The above is a specific embodiment of the present invention. This system makes it possible to detect fraudulent use with high accuracy in real time and to provide an environment in which users can respond quickly.

[1037] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1038] Step 1:

[1039] The server collects credit card transaction data and shipping information. Specifically, it obtains detailed information about each transaction (transaction date and time, amount, store location, shipping information, etc.) through a data collection API. The input is the detailed information about each transaction, and the output is a list of the collected transaction data and shipping information.

[1040] Step 2:

[1041] The server normalizes the collected transaction data and shipping destination information and removes outliers. Specifically, it performs a data cleansing process to remove outliers and missing values ​​and format the data into a unified format. The input is a list of collected transaction data, and the output is the normalized and cleansed data.

[1042] Step 3:

[1043] The server extracts features from the normalized and cleansed data. Specifically, it calculates and extracts features for each transaction by amount, date, time, and region. The input is the normalized and cleansed data, and the output is the data with extracted features.

[1044] Step 4:

[1045] The server trains a generative AI model using the data from which features have been extracted. Specifically, it performs model training to learn normal transaction patterns and fraudulent usage patterns based on past transaction data. The input is the data from which features have been extracted, and the output is the trained generative AI model.

[1046] Step 5:

[1047] When a new transaction occurs, the user's device sends the transaction data to the server in real time. Specifically, the data transmission API is called the moment the transaction is completed. The input is the new transaction data, and the output is the transaction data sent to the server.

[1048] Step 6:

[1049] The server analyzes the transaction data received in real time using a generative artificial intelligence model to detect anomalies. Specifically, the transaction data is input into the model, and the results are judged to be abnormal. The input is the transaction data received in real time and the trained model, and the output is the judgment result of whether or not there is an abnormality.

[1050] Step 7:

[1051] If an anomaly is detected, the server promptly sends a push notification to the user's device. Specifically, it uses the notification service API to send the details of the anomaly detection to the user's device. The input is the detected anomaly information, and the output is the push notification sent to the user.

[1052] Step 8:

[1053] The user receives a push notification on their device and verifies the legitimacy of the transaction. If they determine that the transaction is fraudulent, they can proceed with suspending their credit card through the application. Specifically, by pressing a button within the application, fraudulent use is reported and the card is suspended. The input is the user's confirmation result, and the output is the completion of the card suspension procedure.

[1054] Step 9:

[1055] The server periodically updates transaction data and delivery destination information, and retrains the generative AI model to improve its accuracy. Specifically, it periodically updates data and trains it to learn new fraudulent usage patterns. The input is the latest transaction data, and the output is the updated generative AI model.

[1056] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1057] This invention is a system that collects credit card transaction data and shipping destination information, and uses a generative AI model combined with an emotion engine to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[1058] Data collection and preprocessing

[1059] The server collects credit card transaction data and shipping information for each transaction, as well as known fraud data. The collected data is normalized and outliers are removed. This extracts features and prepares the data for processing.

[1060] Examples:

[1061] When a user purchases a product from an online shop, transaction data (purchase date and time, amount, shop location, etc.) is collected. At the same time, delivery address information for the product is also obtained.

[1062] Model training and generation

[1063] The server uses the preprocessed data to train a generative AI model. This model learns normal usage patterns and the characteristics of fraudulent usage, and is capable of detecting anomalies with high accuracy. Meanwhile, the emotion engine recognizes emotions from the user's voice, facial expressions, and text input, and analyzes the data. Emotional data is also used to train the generative AI model.

[1064] Examples:

[1065] The server analyzes past transaction data to learn, for example, which region a user typically shopped in and within what price range. This allows the model to memorize typical usage patterns. At the same time, it also analyzes user emotional data and incorporates it into the model.

[1066] Real-time detection

[1067] When a user purchases a product, the device sends the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an abnormality is detected, an alert is generated immediately. The user's reaction is also analyzed by an emotion engine, and the user's level of stress and anxiety is also detected.

[1068] Examples:

[1069] If a user makes a sudden, expensive purchase in a specific major city while traveling abroad, the transaction data is sent to the server, and the model determines that it differs from normal patterns. As a result, an anomaly is detected, and analysis by the emotion engine detects the user's feelings of surprise or anxiety.

[1070] User notification and response

[1071] If an abnormality is detected, the server will promptly notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user will go through the necessary procedures to suspend the card or change the shipping address. The emotion engine analyzes the user's reaction and customizes the notification content and response procedures as needed.

[1072] Examples:

[1073] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and identifies that the transaction was not initiated by them, the emotion engine detects that the user is extremely anxious and provides gentle instructions on how to respond. The user can also press a button in the app to immediately report the incident to their credit card company and take steps to suspend their card.

[1074] Continuous model updates

[1075] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy. This allows the system to always perform highly accurate fraud detection based on the latest data. Furthermore, emotion data from an emotion engine is added, enabling more personalized responses.

[1076] Examples:

[1077] The server learns new fraud patterns through monthly data updates and retrains the model, ensuring the system can respond to the latest fraud techniques. Data from the emotion engine is also incorporated into the model, making it possible to customize notification methods and countermeasures based on the user's emotional state.

[1078] In this way, the system of the present invention detects fraudulent credit card use and fraudulent changes to shipping addresses in real time with high accuracy, and provides prompt notification and response that takes into account the user's emotional state, thereby realizing a safe and secure transaction environment for users and reducing their mental burden.

[1079] The processing flow will be explained below.

[1080] Step 1:

[1081] The server periodically collects credit card transaction data and shipping information.

[1082] Specific operation: Using the APIs of credit card companies and delivery companies, the system obtains the latest transaction data and delivery address information, while also collecting data on known past fraudulent use.

[1083] Step 2:

[1084] The server normalizes the collected data, removes outliers, and extracts features.

[1085] Specific operations: Extract key fields such as date, time, amount, store used, and delivery destination from raw data, normalize them into a consistent format, detect and remove outliers, and calculate the features required for analysis.

[1086] Step 3:

[1087] The server trains a generative artificial intelligence model using the preprocessed data.

[1088] Specific operation: The normalized and feature-extracted data is split into a training set and a test set. A generative AI model is trained to learn normal usage patterns and the characteristics of fraudulent usage.

[1089] Step 4:

[1090] When a user makes a transaction using a credit card, the terminal transmits the transaction data to the server in real time.

[1091] Specific operation: The POS system or online payment platform immediately sends the user's transaction data (e.g., time, location, amount, etc.) to the server.

[1092] Step 5:

[1093] The server immediately analyzes the received transaction data using a generative artificial intelligence model and detects any anomalies.

[1094] How it works: Incoming data is fed into a generative AI model in real time, where it is compared with normal usage patterns. If an abnormal pattern is detected, an anomaly flag is raised.

[1095] Step 6:

[1096] If an abnormality is detected, the server will promptly notify the user.

[1097] Specific behavior: If an abnormality flag is raised, a warning message will be sent using the user's contact information (email address, SMS, push notification, etc.).

[1098] Step 7:

[1099] The emotion engine analyzes the user's reaction in real time when they receive a notification and detects their emotions (e.g., surprise, anxiety, anger, etc.).

[1100] Specific operation: The emotion engine analyzes the user's voice, facial expressions, text input, etc. when receiving the notification to determine their emotional state.

[1101] Step 8:

[1102] The user receives a notification and verifies the authenticity of the transaction.

[1103] Specific operation: Upon receiving the notification, the user accesses the web portal or smartphone app to verify whether the transaction was initiated by them.

[1104] Step 9:

[1105] The server customizes the notification content and response procedures based on the user's emotions detected by the emotion engine.

[1106] Specific behavior: If the user's emotions indicate anxiety or surprise, the server will re-notify them with softer language and reassuring messages, and if necessary, instruct them to escalate the issue to a support staff member.

[1107] Step 10:

[1108] If the user determines that the use has been fraudulent, procedures such as suspending the card or changing the delivery address will be implemented.

[1109] Specific operation: If a user determines that a transaction has been fraudulent, they can contact their credit card company using a dedicated app or a link in the web portal to suspend their card. They can also contact the delivery company to suspend or change the delivery of their product.

[1110] Step 11:

[1111] The server periodically updates transaction data and delivery destination information and retrains the generative artificial intelligence model.

[1112] How it works: Regularly prepare new datasets and evaluate the performance of existing generative AI models. Retrain as needed to maintain or improve model accuracy. Additionally, by adding data from the emotion engine, the model can be trained to provide more personalized responses.

[1113] Through these steps, the system of the present invention detects credit card fraud and shipping fraud in real time and enables quick and appropriate responses based on the user's emotional state.

[1114] Example 2

[1115] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1116] In modern society, credit card fraud is on the rise, with the risk increasing especially in online shopping and remote transactions. When fraud occurs, it not only causes financial loss to users, but also causes significant psychological stress. Furthermore, existing fraud detection systems are often inadequate because they struggle to respond in real time and are unable to take into account the user's emotional state. This makes it difficult for many users to conduct transactions with confidence. Therefore, there is a need for a system that can detect fraud quickly and accurately, and provide notifications and responses that take the user's emotions into account.

[1117] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for collecting credit card transaction data and delivery destination information, means for normalizing the transaction data and delivery destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for collecting and analyzing user emotion data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, and means for notifying the user of detected anomalies and allowing the user to confirm the anomaly. This makes it possible to quickly and accurately detect fraudulent credit card use and to notify and respond in consideration of the user's emotional state.

[1118] "Credit card transaction data" means information that records purchases and payments made using a credit card.

[1119] "Delivery destination information" refers to information such as the address and contact details of the recipient of the purchased product or service.

[1120] "Normalization" is the process of standardizing the range and units of data to make it easier to compare and analyze.

[1121] "Outlier removal" is the process of identifying and eliminating invalid data points within a data set.

[1122] "Features" are important input variables used in machine learning models for prediction and classification.

[1123] A "generative artificial intelligence model" is an AI model that has the ability to learn from data, recognize patterns, and generate or analyze new data.

[1124] "User emotion data" refers to information about the user's emotional state obtained from the user's facial expression, voice, text input, and the like.

[1125] "Receiving transaction data in real time" means that the data is sent to the server and received immediately the moment a transaction occurs.

[1126] "Anomaly detection" refers to identifying fraudulent or unusual behavior that deviates from normal patterns.

[1127] "Notification" is the act of the system informing the user of abnormalities or important information.

[1128] This invention relates to a highly accurate and rapid detection system for preventing fraudulent use of credit cards. This system collects credit card transaction data and shipping destination information, and uses this data to learn and apply a generative artificial intelligence model that detects fraudulent use. The details are as follows:

[1129] Data collection and preprocessing

[1130] The server collects credit card transaction data and shipping information for each transaction, which involves retrieving the data through an API and storing it in a database using MySQL, a popular relational database management system (RDBMS).

[1131] Specifically, when a user purchases an item from an online shop, the transaction data (purchase date and time, amount, store location, etc.) and shipping information are stored in a MySQL database. The collected data is read using the Python Pandas library, normalized, and outliers are removed, preparing the data for processing.

[1132] Model training and generation

[1133] The server uses the preprocessed data to train a generative artificial intelligence model. This model learns normal usage patterns and the characteristics of fraudulent usage, and detects anomalies with high accuracy. The specific software used is Scikit-learn's RandomForestClassifier.

[1134] The model training uses transaction data and fraudulent use data collected to date. Additionally, Affectiva's SDK is used as a sentiment analysis library to collect and analyze user sentiment data. This sentiment data is also fed back into the model training.

[1135] Specifically, the server analyzes past transaction data to learn, for example, which region a user typically spends money in. At the same time, it analyzes emotional data using Affectiva's SDK and incorporates that data into the model.

[1136] Real-time detection

[1137] When a user purchases a product, the device sends the transaction data to the server in real time. The server immediately analyzes the received data using a generative artificial intelligence model and immediately generates a warning if an abnormality is detected. The server also analyzes the user's reactions using an emotion engine to detect the user's level of stress and anxiety.

[1138] Specifically, if a user makes a sudden, expensive purchase in a specific major city while traveling abroad, the transaction data is sent to the server, and the model determines that the transaction is different from the usual pattern. If it determines that fraudulent use is suspected, emotion analysis can also detect the user's anxiety.

[1139] User notification and response

[1140] The server uses Firebase Cloud Messaging as a communication library to promptly notify the user if an abnormality is detected. The user who receives the notification can check whether the transaction is legitimate, and if it is fraudulent, can immediately take action such as suspending the card or changing the delivery address.

[1141] Specifically, when the server detects an abnormality, it sends a push notification to the user's smartphone. If the user checks the notification and determines that the transaction was not initiated by them, the app uses emotional analysis to detect that the user is extremely anxious and provides gentle guidance. The app also allows the user to press a button to report the transaction to their credit card company and immediately take steps to suspend their card.

[1142] Continuous model updates

[1143] The server periodically updates transaction data and delivery destination information, and performs re-training to improve the accuracy of the generative AI model. This ensures highly accurate fraud detection based on the latest data. Emotional data is also added, enabling more personalized responses.

[1144] Specifically, the server updates data monthly, learns new fraudulent usage patterns, and retrains the model. This ensures the system can respond to the latest fraud techniques. Emotional data is also fed back to the model, allowing it to provide notification methods and countermeasures tailored to the user.

[1145] Prompt Sentence Examples

[1146] Fraud detection is performed on newly collected transaction data, including:

[1147] Purchase date: 2023-10-01 10:00:00

[1148] Amount: $500

[1149] Shop location: London

[1150] Delivery Address: 123 ABC Street, London

[1151] User emotion data: Moderate anxiety

[1152] Use this data to detect anomalies in real time using fraud detection models and notify users as needed.

[1153] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1154] System program processing flow

[1155] Step 1:

[1156] The server collects credit card transaction data and shipping information. Specifically, it uses an API to retrieve credit card transaction information and stores it in a MySQL database. The input includes the transaction data and shipping information retrieved from the API. The output is this raw data stored in the database.

[1157] Specific behavior:

[1158] Make API calls to retrieve transaction data and shipping information.

[1159] The retrieved data is stored in a MySQL database.

[1160] Step 2:

[1161] The server pre-processes the collected data. It uses the Pandas library to normalize the data, remove outliers, and extract features. The input includes raw data stored in a MySQL database. The output is the pre-processed, feature-extracted data.

[1162] Specific behavior:

[1163] Use Pandas to read data from a MySQL database.

[1164] Normalize the data to make it consistent across ranges.

[1165] Detect and remove outliers.

[1166] Important features (e.g., purchase date and time, amount, store location) are extracted and saved as structured data.

[1167] Step 3:

[1168] The server uses the preprocessed data to train a generative AI model. It uses Scikit-learn's RandomForestClassifier to learn normal transaction patterns and fraudulent usage patterns. The input includes training data with extracted features. The output is a trained generative AI model.

[1169] Specific behavior:

[1170] Define a model using Scikit-learn's RandomForestClassifier.

[1171] The data from which the features have been extracted is supplied to the model as training data and fitted.

[1172] The model is evaluated and parameters are adjusted as necessary.

[1173] Step 4:

[1174] The server collects and analyzes the user's emotional data. Using Affectiva's SDK, it analyzes the user's emotions (e.g., stress, anxiety, surprise, etc.) and stores the data. Inputs include the user's facial expressions, voice, and text input. The output is the analyzed emotional data.

[1175] Specific behavior:

[1176] Affectiva's SDK is used to analyze the user's facial expressions, voice, and text input.

[1177] The analysis results (emotion data) are saved as structured data and provided to a generative artificial intelligence model.

[1178] Step 5:

[1179] When a user purchases a product, the device sends transaction data to a server in real time. The server analyzes the received data using a generative artificial intelligence model to detect anomalies. The input includes real-time transaction data and a pre-trained model. The output is a warning message or notification when an anomaly is detected.

[1180] Specific behavior:

[1181] The terminal collects transaction data and immediately transmits it to the server.

[1182] The server receives the transaction data and inputs it into a generative artificial intelligence model.

[1183] If the model detects an anomaly, it generates an anomaly alert.

[1184] Step 6:

[1185] The server notifies the user of the detected anomaly, and the user confirms the anomaly. The push notification is sent using Firebase Cloud Messaging. The input includes the transaction data and sentiment data for which the anomaly was detected. The output is the notification and confirmation for the user.

[1186] Specific behavior:

[1187] Use Firebase Cloud Messaging to send push notifications to users' devices.

[1188] Users can review notifications and approve or report unusual transactions within the app.

[1189] Step 7:

[1190] The server periodically updates the transaction data and shipping destination information and retrains the generative AI model to improve its accuracy. The input includes new transaction data and updated emotion data. The output is a retrained, highly accurate generative AI model.

[1191] Specific behavior:

[1192] Extract new data from a MySQL database.

[1193] Retrain the model and reevaluate its performance.

[1194] Adjust parameters as needed to improve the accuracy of the model.

[1195] (Application example 2)

[1196] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1197] Fraudulent credit card use is becoming more sophisticated every year, making it difficult to quickly and accurately detect fraudulent transactions using conventional methods. Furthermore, when a fraudulent transaction is detected, notifications are given to users uniformly, and no response is given based on the user's emotional state. This often leaves users feeling anxious and stressed. The present invention aims to solve these problems and realize effective fraud detection and smooth user response.

[1198] The identification processing by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting credit card transaction data and delivery destination information, means for normalizing the transaction data and delivery destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for collecting and analyzing user emotion data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, and means for notifying the user of detected anomalies, customizing the content of the notification based on the user's emotional state, and allowing the user to confirm the anomaly. This enables highly accurate real-time detection of fraudulent transactions and enables appropriate notification and response tailored to the user's emotional state.

[1199] Output for definition statements

[1200] "Credit card transaction data" refers to information about transactions conducted using a credit card, including details such as the transaction date and time, transaction amount, and transaction store.

[1201] "Shipping destination information" refers to information about the shipping destination of the purchased product, specifically information such as address, name, and contact details.

[1202] "Normalization" refers to the process of maintaining data consistency and ensuring consistency across different data sets.

[1203] An "outlier" is a data point that falls outside the normal range, such as an extremely high or low value in a data set.

[1204] "Features" refer to specific attributes or parameters of input data that a generative artificial intelligence model uses for learning and prediction.

[1205] A "generative artificial intelligence model" is a statistical or machine learning model that learns patterns and features based on large amounts of data and makes predictions and classifications.

[1206] "Emotional data" refers to information about the user's emotional state obtained from their voice, facial expression, text input, etc.

[1207] "Real-time" refers to immediate processing or reaction, meaning data is collected, analyzed, and responded to with short latency.

[1208] "Anomaly detection" refers to the process of automatically identifying fraudulent activity or anomalous behavior that deviates from normal usage patterns.

[1209] "Customizing notification content" refers to changing the message and method of notification to a user to suit the status and characteristics of each individual user.

[1210] MODE FOR CARRYING OUT THE INVENTION

[1211] Overall system overview

[1212] The system of the present invention includes components for detecting fraudulent credit card use with high accuracy and for providing prompt and appropriate responses to users. The system is primarily composed of a server, terminals, and users.

[1213] Server Processing

[1214] The server processes the data in the following steps to detect and notify anomalies.

[1215] 1. Data Collection

[1216] The server collects credit card transaction data and shipping information, including details such as transaction date and time, amount, currency, merchant, and shipping address.

[1217] Hardware: The servers are general-purpose servers equipped with high-performance processors and storage.

[1218] Software: Data is collected through a database management system (DBMS) or API.

[1219] 2. Data Preprocessing

[1220] The collected data is normalized, outliers are removed, and features are extracted, providing data suitable for training generative AI models.

[1221] Software used: Data preprocessing scripts using programming languages ​​such as Python and R, and data manipulation libraries such as Pandas and NumPy.

[1222] 3. Learning generative AI models

[1223] The normalized and feature-extracted data is used to train a generative AI model, which learns the characteristics of normal and fraudulent usage patterns.

[1224] Hardware used: High-performance server equipped with GPU etc.

[1225] Software used: Machine learning libraries such as TensorFlow and PyTorch.

[1226] 4. Emotional Data Collection and Analysis

[1227] It collects and analyzes emotional data such as voice and facial expressions from users' smartphones and other devices, uses an emotion engine to identify emotional states, and then applies that data to generative AI models.

[1228] Hardware used: Smartphone camera and microphone.

[1229] Software used: Speech recognition API and image analysis API.

[1230] 5. Real-time detection and notification

[1231] It receives real-time data as transactions occur, analyzes it with a generative AI model, and sends customized notifications based on the user's emotional state if an anomaly is detected.

[1232] Hardware used: High-performance server capable of real-time processing.

[1233] Software used: Real-time data streaming framework, push notification service.

[1234] Specific examples

[1235] When a user makes a transaction using a credit card, the data (e.g., the transaction amount is 150,000 yen, and the transaction location is New York) is sent to the server. The server immediately analyzes this transaction data using a generative AI model and compares it with normal usage patterns. If the transaction is determined to be abnormal, the server uses an emotion engine to obtain emotional data from the smartphone's camera and microphone and detects that the user is anxious. As a result, it can send a notification adapted to the user's emotional state, such as displaying a message saying, "There is a suspicion of a fraudulent transaction. Don't worry, we will take immediate action."

[1236] Prompt Sentence Examples

[1237] Here are some examples of prompts for generative AI models:

[1238] Please determine whether the following credit card transaction is legitimate. Transaction information: Amount: 150,000 yen, Location: New York, User: Tokyo. Is this transaction different from the usual pattern? Also, the user's voice indicates anxiety.

[1239] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1240] Program processing steps

[1241] Step 1:

[1242] The server collects all credit card transaction data and shipping information. For each transaction, details such as transaction date and time, transaction amount, currency, transaction store, shipping address, etc. are stored in a database. This input data is recorded as a record for each transaction.

[1243] Step 2:

[1244] The server normalizes the collected transaction data and removes outliers. This includes maintaining data consistency and standardizing the format. For example, it detects and removes transactions with unusually high prices or inaccurate shipping information. This process produces clean, consistent feature data.

[1245] Step 3:

[1246] Using the normalized and feature-extracted data, the server trains a generative AI model. This model learns past patterns of normal and fraudulent usage. The data inputs are historical transaction data and its features, and the output is feedback that the model uses to classify normal and abnormal behavior.

[1247] Step 4:

[1248] A user's smartphone or other device collects and sends emotional data to a server. Emotional data includes information such as the user's voice and facial expressions. The emotion engine analyzes this data and identifies the user's emotional state (e.g., anxiety, stress). This data is the input data obtained through the emotion engine API, and the analyzed emotional state is the output.

[1249] Step 5:

[1250] The server receives transaction data in real time and analyzes it using a generative AI model. Here, detailed transaction information is the input data, and the output is an anomaly detected as the analysis result of the AI ​​model. If the AI ​​detects an anomaly, the details are immediately passed on to the next step.

[1251] Step 6:

[1252] The server notifies the user of detected anomalies and customizes the notification content based on the user's emotional state. The notification message is generated based on the user's emotional state obtained by the emotion engine and the transaction data. For example, if the server detects that the user is feeling anxious or stressed, the notification content is customized with gentler language such as "Don't worry, we will take immediate action." The output data of the emotion engine and the abnormal transaction data are used as input, and the output is a notification message to the user.

[1253] Step 7:

[1254] The user receives a notification through their device and checks whether the abnormal transaction is legitimate. If the user checks the transaction and determines it to be fraudulent, they can send instructions to the server via the app, such as suspending their credit card or changing the shipping address. This inputs the user's confirmation data, and the terminal then sends instructions to the card company, etc., which is the output.

[1255] Step 8:

[1256] The server periodically updates transaction data and delivery destination information, retraining the generative AI model to maintain or improve its accuracy. Emotional data is also periodically added and incorporated into the model. This allows the system to always use the latest information for highly accurate fraud detection. The latest transaction data and emotional data are used as input, and the output is an updated AI model.

[1257] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1258] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1259] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1260] [Fourth embodiment]

[1261] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1262] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1263] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1264] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1265] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1266] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1267] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1268] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1269] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1270] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1271] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1272] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1273] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1274] This invention is a system that collects credit card transaction data and shipping destination information, and uses a generative artificial intelligence model to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[1275] Data collection and preprocessing

[1276] The server collects credit card transaction data and shipping information for each transaction, as well as known fraud data. The collected data is normalized and outliers are removed. This extracts features and prepares the data for processing.

[1277] Examples:

[1278] When a user purchases a product from an online shop, transaction data (purchase date and time, amount, shop location, etc.) is collected. At the same time, delivery address information for the product is also obtained.

[1279] Model training and generation

[1280] The server uses the preprocessed data to train a generative artificial intelligence model, which learns normal usage patterns and the characteristics of fraudulent usage, and has the ability to detect anomalies with high accuracy.

[1281] Examples:

[1282] The server analyzes past transaction data to learn, for example, which region a user typically spends money in. This allows the model to memorize typical usage patterns.

[1283] Real-time detection

[1284] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, an alert is generated immediately.

[1285] Examples:

[1286] If a user makes a sudden, expensive purchase in a particular major city while traveling abroad, the transaction data is sent to the server, and the model determines that it differs from the normal pattern, detecting the anomaly and generating a warning.

[1287] User notification and response

[1288] If the server detects an abnormality, it will immediately notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user will take steps to suspend the card or change the shipping address.

[1289] Examples:

[1290] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and, for example, determines that the transaction was not initiated by them, they can press a button in the app to immediately report the incident to their credit card company and take steps to suspend their card.

[1291] Continuous model updates

[1292] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy, allowing the system to always perform highly accurate fraud detection based on the latest data.

[1293] Examples:

[1294] Through monthly data updates, the server learns new fraud patterns and retrains the model, keeping the system up to date with the latest fraud techniques.

[1295] In this way, the system of the present invention detects fraudulent credit card use and fraudulent changes to shipping addresses in real time with high accuracy, and provides prompt notification and response to users, thereby realizing a safe and secure transaction environment for users.

[1296] The processing flow will be explained below.

[1297] Step 1:

[1298] The server periodically collects credit card transaction data and shipping information.

[1299] Specific operation: Using the APIs of credit card companies and delivery companies, the system obtains the latest transaction data and delivery address information, while also collecting data on known past fraudulent use.

[1300] Step 2:

[1301] The server normalizes the collected data, removes outliers, and extracts features.

[1302] Specific operations: Extracts key fields such as date, time, amount, store used, and delivery destination from raw data and normalizes them into a consistent format. Detects and removes outliers and calculates the features required for analysis.

[1303] Step 3:

[1304] The server trains a generative artificial intelligence model using the preprocessed data.

[1305] Specific operations: Split the normalized and feature-extracted data into a training set and a test set, and train a generative AI model to learn normal usage patterns and the characteristics of fraudulent usage.

[1306] Step 4:

[1307] When a user makes a transaction using a credit card, the terminal transmits the transaction data to the server in real time.

[1308] What it does: The POS system or online payment platform instantly sends the user's transaction data (e.g., time, location, amount, etc.) to the server.

[1309] Step 5:

[1310] The server immediately analyzes the received transaction data using a generative artificial intelligence model and detects any anomalies.

[1311] How it works: Incoming data is fed into a generative AI model in real time, where it is compared with normal usage patterns. If an abnormal pattern is detected, an anomaly flag is raised.

[1312] Step 6:

[1313] If an abnormality is detected, the server will promptly notify the user.

[1314] Specific behavior: If an anomaly is flagged, a warning message will be sent using the user's contact information (email address, SMS, push notification, etc.).

[1315] Step 7:

[1316] The user receives a notification and verifies the authenticity of the transaction.

[1317] Specific operation: When a user receives a notification, they access a web portal or smartphone app to verify whether the transaction was initiated by them.

[1318] Step 8:

[1319] If the user determines that the use has been fraudulent, procedures such as suspending the card or changing the delivery address will be implemented.

[1320] Specific actions: If a user determines that a transaction has been fraudulent, they can contact their credit card company using a link in the dedicated app or web portal to suspend their card. They can also contact the delivery company to suspend or change the delivery of their product.

[1321] Step 9:

[1322] The server periodically updates transaction data and delivery destination information and retrains the generative artificial intelligence model.

[1323] What it does: Periodically prepare new datasets to evaluate the performance of existing generative AI models, retraining them as needed to maintain or improve their accuracy.

[1324] Through the above steps, the system of the present invention detects credit card fraud and shipping address fraud in real time, promptly notifying the user so that appropriate action can be taken.

[1325] Example 1

[1326] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1327] Fraudulent credit card use can cause serious financial damage to users. Early detection of fraudulent use and appropriate countermeasures are required, but current systems have difficulty detecting fraud in real time with high accuracy. Furthermore, maintaining the accuracy of the system requires continuous model updates, which are difficult to achieve efficiently with conventional methods.

[1328] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1329] In this invention, the server includes means for collecting credit card transaction data and recipient information, means for normalizing the transaction data and recipient information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, means for notifying the user of detected anomalies, allowing the user to confirm the anomaly and, if fraudulent, suspending the payment method or changing the recipient information, and means for periodically updating the transaction data and recipient information and relearning the generative artificial intelligence model to maintain or improve its accuracy. This makes it possible to detect fraudulent credit card use in real time with high accuracy and provide users with prompt notification and support.

[1330] A "credit card" is a payment method issued by a bank or credit card company that users use when purchasing goods or services.

[1331] "Transaction data" refers to information relating to purchases or payments made using a credit card, including, for example, the date and time of purchase, the amount, and the place of purchase.

[1332] "Recipient information" refers to information about the delivery address of purchased products or services, and is data including information such as the address, recipient name, and contact details.

[1333] "Normalization" is the process of converting data values ​​to a uniform scale to ensure data consistency and comparability.

[1334] An "outlier" is a value that deviates from the normal data pattern and should be treated as noise or an error in data analysis.

[1335] In data analysis and machine learning, a "feature" is a numerical value or indicator that represents an important attribute or pattern of data.

[1336] A "generative artificial intelligence model" is a type of artificial intelligence that is an algorithm that learns patterns and characteristics of data and makes predictions and detects anomalies in new data.

[1337] "Real-time" refers to processing data either instantly as it occurs or with very little delay.

[1338] "Anomaly detection" is the process of identifying data that deviates from normal patterns, indicating potential problems or fraud.

[1339] "User" means an end user who purchases goods or services using a credit card and receives notifications from the system.

[1340] "Payment Instrument" means a method of paying for goods and services, including credit cards and debit cards.

[1341] "Model retraining" is the process of using new collected data to readjust the model parameters and algorithms in order to maintain or improve the accuracy of a generative artificial intelligence model.

[1342] This invention is a system that collects credit card transaction data and recipient information, and uses a generative artificial intelligence model to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[1343] Data collection and preprocessing

[1344] The server collects credit card transaction data and payee information for each transaction, as well as known fraud data. Through an API, the server normalizes the collected data and removes outliers. This extracts features and prepares the data for processing.

[1345] Examples:

[1346] For example, when a user purchases a product from an online shop, the transaction data (purchase date and time, amount, shop location, etc.) is sent to the server. At the same time, the server also collects information about the recipient of the product.

[1347] Model learning

[1348] The server uses the preprocessed data to train a generative AI model, which then learns normal usage patterns and the characteristics of fraudulent usage, detecting anomalies with high accuracy.

[1349] Examples:

[1350] The server analyzes past transaction data to learn, for example, which region a user typically spends money in. This allows the server to memorize typical usage patterns into a model.

[1351] Real-time detection

[1352] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, an alert is generated immediately.

[1353] Examples:

[1354] For example, if a user makes a large purchase in a particular major city while traveling abroad, the transaction data is sent to the server, and the model determines that the transaction differs from the normal pattern. As a result, an anomaly is detected and an alert is generated.

[1355] User notification and response

[1356] If the server detects an abnormality, it will immediately notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user can take steps to stop the payment method or change the payee information.

[1357] Examples:

[1358] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and, for example, determines that the transaction was not made by them, they can press a button in the app to immediately report it to their credit card company and take steps to suspend their credit card.

[1359] Continuous model updates

[1360] The server periodically updates transaction data and recipient information, and retrains the generative AI model to maintain or improve its accuracy, allowing the system to always perform highly accurate fraud detection based on the latest data.

[1361] Examples:

[1362] Every month, the server retrains the model using a new dataset collected, updating the model parameters to the latest state, thereby keeping the system adaptable to new fraud techniques.

[1363] Prompt Sentence Examples

[1364] "Please retrain your fraud detection model using this month's transaction data."

[1365] This system makes it possible to detect fraudulent credit card use in real time with high accuracy, and to provide users with prompt notification and response.

[1366] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1367] Step 1:

[1368] Data collection

[1369] The server collects transaction data and recipient information from the APIs of credit card companies and online shops. Inputs include the purchase date and time, amount, shop location, and recipient information for each transaction. This data is stored on the server.

[1370] Specific behavior:

[1371] When a new transaction occurs, the server receives the transaction data in real time through the API. For example, when a user purchases a product, the transaction information is sent to the server.

[1372] Step 2:

[1373] Fraud data collection

[1374] The server also collects known fraud data. The input is transaction data of previously detected fraudulent activity. This data serves as the basis for learning fraud patterns.

[1375] Specific behavior:

[1376] The server periodically downloads fraud data from credit card companies and other data providers.

[1377] Step 3:

[1378] Data normalization and outlier removal

[1379] The server normalizes the collected transaction data and recipient information and removes outliers. The input is the collected raw data, and the output is the normalized data. This keeps the data consistent and makes it easier to analyze.

[1380] Specific behavior:

[1381] Check the transaction dataset stored in the database and perform preprocessing such as scaling and missing value imputation.

[1382] Step 4:

[1383] Feature extraction

[1384] The server extracts features from the preprocessed data. The input is normalized data, and the output is a set of features. The features include transaction frequency, amount statistics, geographic information, etc.

[1385] Specific behavior:

[1386] For each transaction, multiple metrics (purchase date and time, average and variance of the amount, geographical distance, etc.) are calculated to create a feature set.

[1387] Step 5:

[1388] Learning generative artificial intelligence models

[1389] The server uses the extracted features to train a generative AI model. The input is a set of features, and the output is a trained generative AI model. This allows the server to learn the characteristics of normal usage patterns and fraudulent usage.

[1390] Specific behavior:

[1391] Historical trading data is used as a training set to optimize the model parameters.

[1392] Step 6:

[1393] Real-time data transmission

[1394] When a user purchases a product, the terminal transmits the transaction data to the server in real time. The input is the user's transaction data, and the transmitted data arrives at the server.

[1395] Specific behavior:

[1396] The user's purchase information is immediately sent from the terminal to the server.

[1397] Step 7:

[1398] Real-time analytics

[1399] The server instantly analyzes the received transaction data using a generative artificial intelligence model. The input is the transaction data sent in real time, and the output is the presence or absence of anomalies. If there is a possibility of fraud, a warning is generated.

[1400] Specific behavior:

[1401] The data is analyzed layer by layer, and when an anomaly is detected, an anomaly score is generated.

[1402] Step 8:

[1403] User Notifications

[1404] If an abnormality is detected, the server will promptly send a notification to the user. The input is the abnormality detection result, and the output is a notification message to the user.

[1405] Specific behavior:

[1406] When an anomaly is detected, the server generates a notification message and sends a push notification to the user's smartphone.

[1407] Step 9:

[1408] User Support

[1409] The user receives a notification and verifies whether the transaction is legitimate. If it is fraudulent, they can block the payment method or change the payee information. The input is the user's confirmation, and the output is an action such as reporting to the card company.

[1410] Specific behavior:

[1411] If the user checks the notification and presses the "This transaction was not made by me" button, the transaction will be reported to the credit card company and the credit card will be suspended.

[1412] Step 10:

[1413] Model Update

[1414] The server periodically updates the transaction data and recipient information and retrains the generative AI model to maintain or improve its accuracy. The input is a new dataset and the output is an updated generative AI model.

[1415] Specific behavior:

[1416] Every month, the model is retrained using the latest collected data and the model parameters are optimized.

[1417] (Application example 1)

[1418] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1419] Conventional credit card fraud detection systems were slow to detect fraud, with users often only realizing they had been compromised after the fraud had actually occurred. Furthermore, they lacked the functionality to allow users to respond quickly after fraud was detected, making it difficult to prevent fraudulent use. Furthermore, as fraud patterns change, continuous system updates are required, which was difficult to achieve with conventional systems.

[1420] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1421] In this invention, the server includes means for collecting credit card transaction data and shipping destination information, means for normalizing the transaction data and shipping destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for receiving the transaction data in real time and detecting anomalies using the generative artificial intelligence model, means for sending a push notification of the detected anomaly to the user's communication terminal, and means for allowing the user to immediately take steps to suspend their credit card based on the results of the generative artificial intelligence model. This enables highly accurate detection of fraudulent use in real time and provides an environment in which users can respond quickly.

[1422] A "credit card" is a payment method with a certain credit limit that is used when purchasing goods and services.

[1423] "Transaction Data" refers to detailed information about purchases made using a credit card, including transaction date and time, amount, store location, and other data.

[1424] "Shipping destination information" refers to information about the shipping destination of products purchased with a credit card, and includes the address and name of the recipient.

[1425] "Normalization" is the process of converting data into a consistent format, a technique that makes data analysis and model training easier by standardizing the scale of variables.

[1426] An "outlier" is a piece of data that is statistically significantly different from other data points, and is a number that may affect the results of an analysis.

[1427] "Features" are input data provided to a machine learning model that represent characteristics or attributes that are important for the model to learn data patterns.

[1428] A "generative artificial intelligence model" is a machine learning algorithm designed to learn the patterns and rules needed to perform a specific task, and is typically powered by large amounts of training data.

[1429] "Learning" is the process by which a generative artificial intelligence model finds patterns and rules in given data and improves its performance.

[1430] "Real-time" refers to a method in which data is processed as it is acquired and results are provided immediately.

[1431] "Anomaly detection" is a function that finds behaviors or values ​​in data that differ from normal patterns, and is used to identify fraudulent use or abnormal behavior.

[1432] A "push notification" is a notification message that an application sends directly to a user's communication terminal, and is a means of delivering information to the user immediately.

[1433] "Credit card suspension procedure" is a process to temporarily or permanently disable the use of a credit card suspected of fraudulent use, and is carried out to prevent further damage.

[1434] The system for realizing this invention collects credit card transaction data and shipping destination information, detects fraudulent use in real time using a generative artificial intelligence model, and notifies users. This system is composed of a server, terminals, and users.

[1435] Data collection and preprocessing

[1436] The server collects credit card transaction data and shipping address information for each transaction. The collected data is normalized and outliers are removed. This allows features to be extracted and the data is ready for processing. For example, when a user purchases a product online, the transaction data (purchase date and time, amount, purchase location, etc.) is collected. At the same time, the product's shipping address information is also obtained.

[1437] Model learning

[1438] The server uses the preprocessed data to train a generative artificial intelligence model. This model learns the characteristics of normal usage patterns and fraudulent usage, and has the ability to detect anomalies with high accuracy. For example, the server analyzes past transaction data to learn which region a user is shopping in and within what price range. This allows normal usage patterns to be memorized in the model.

[1439] Real-time detection

[1440] When a user purchases a product, the user's device sends the transaction data to a server in real time. The data received by the server is immediately analyzed by a generative artificial intelligence model, and if an anomaly is detected, a warning is immediately generated. For example, if a user makes an expensive purchase in an unusual location while traveling abroad, the transaction data is sent to the server and is determined to be abnormal by the model. As a result, the anomaly is immediately detected and a warning is generated.

[1441] User notification and response

[1442] If an abnormality is detected, the server will promptly send a notification to the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user can take steps to suspend their credit card. For example, when the server detects an abnormality, a push notification will be sent to the user's smartphone. If the user checks the notification and determines that the transaction was not made by them, they can press a button in the app to report it to their credit card company and take steps to suspend their card.

[1443] Continuous model updates

[1444] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy. This allows the system to always perform highly accurate fraud detection based on the latest data. For example, the server updates data monthly, learns new fraudulent usage patterns, and retrains the model. This makes it possible to respond to the latest fraud techniques.

[1445] Prompt Sentence Examples

[1446] Examples of prompts that can be provided to generative AI models include:

[1447] "Does the purchase of over 5,000 yen in Shinjuku Ward match a pattern that has not been identified as abnormal in your past transaction history?"

[1448] The above is a specific embodiment of the present invention. This system makes it possible to detect fraudulent use with high accuracy in real time and to provide an environment in which users can respond quickly.

[1449] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1450] Step 1:

[1451] The server collects credit card transaction data and shipping information. Specifically, it obtains detailed information about each transaction (transaction date and time, amount, store location, shipping information, etc.) through a data collection API. The input is the detailed information about each transaction, and the output is a list of the collected transaction data and shipping information.

[1452] Step 2:

[1453] The server normalizes the collected transaction data and shipping destination information and removes outliers. Specifically, it performs a data cleansing process to remove outliers and missing values ​​and format the data into a unified format. The input is a list of collected transaction data, and the output is the normalized and cleansed data.

[1454] Step 3:

[1455] The server extracts features from the normalized and cleansed data. Specifically, it calculates and extracts features for each transaction by amount, date, time, and region. The input is the normalized and cleansed data, and the output is the data with extracted features.

[1456] Step 4:

[1457] The server trains a generative AI model using the data from which features have been extracted. Specifically, it performs model training to learn normal transaction patterns and fraudulent usage patterns based on past transaction data. The input is the data from which features have been extracted, and the output is the trained generative AI model.

[1458] Step 5:

[1459] When a new transaction occurs, the user's device sends the transaction data to the server in real time. Specifically, the data transmission API is called the moment the transaction is completed. The input is the new transaction data, and the output is the transaction data sent to the server.

[1460] Step 6:

[1461] The server analyzes the transaction data received in real time using a generative artificial intelligence model to detect anomalies. Specifically, the transaction data is input into the model, and the results are judged to be abnormal. The input is the transaction data received in real time and the trained model, and the output is the judgment result of whether or not there is an abnormality.

[1462] Step 7:

[1463] If an anomaly is detected, the server promptly sends a push notification to the user's device. Specifically, it uses the notification service API to send the details of the anomaly detection to the user's device. The input is the detected anomaly information, and the output is the push notification sent to the user.

[1464] Step 8:

[1465] The user receives a push notification on their device and verifies the legitimacy of the transaction. If they determine that the transaction is fraudulent, they can proceed with suspending their credit card through the application. Specifically, by pressing a button within the application, fraudulent use is reported and the card is suspended. The input is the user's confirmation result, and the output is the completion of the card suspension procedure.

[1466] Step 9:

[1467] The server periodically updates transaction data and delivery destination information, and retrains the generative AI model to improve its accuracy. Specifically, it periodically updates data and trains it to learn new fraudulent usage patterns. The input is the latest transaction data, and the output is the updated generative AI model.

[1468] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1469] This invention is a system that collects credit card transaction data and shipping destination information, and uses a generative AI model combined with an emotion engine to detect fraudulent use in real time. This system consists of a server, terminals, and users.

[1470] Data collection and preprocessing

[1471] The server collects credit card transaction data and shipping information for each transaction, as well as known fraud data. The collected data is normalized and outliers are removed. This extracts features and prepares the data for processing.

[1472] Examples:

[1473] When a user purchases a product from an online shop, transaction data (purchase date and time, amount, shop location, etc.) is collected. At the same time, delivery address information for the product is also obtained.

[1474] Model training and generation

[1475] The server uses the preprocessed data to train a generative AI model. This model learns normal usage patterns and the characteristics of fraudulent usage, and is capable of detecting anomalies with high accuracy. Meanwhile, the emotion engine recognizes emotions from the user's voice, facial expressions, and text input, and analyzes the data. Emotional data is also used to train the generative AI model.

[1476] Examples:

[1477] The server analyzes past transaction data to learn, for example, which region a user typically shopped in and within what price range. This allows the model to memorize typical usage patterns. At the same time, it also analyzes user emotional data and incorporates it into the model.

[1478] Real-time detection

[1479] When a user purchases a product, the device sends the transaction data to the server in real time. The received data is immediately analyzed by a generative artificial intelligence model, and if an abnormality is detected, an alert is generated immediately. The user's reaction is also analyzed by an emotion engine, and the user's level of stress and anxiety is also detected.

[1480] Examples:

[1481] If a user makes a sudden, expensive purchase in a specific major city while traveling abroad, the transaction data is sent to the server, and the model determines that it differs from normal patterns. As a result, an anomaly is detected, and analysis by the emotion engine detects the user's feelings of surprise or anxiety.

[1482] User notification and response

[1483] If an abnormality is detected, the server will promptly notify the user. The user will receive the notification and check whether the transaction is legitimate. If it is fraudulent, the user will go through the necessary procedures to suspend the card or change the shipping address. The emotion engine analyzes the user's reaction and customizes the notification content and response procedures as needed.

[1484] Examples:

[1485] When the server detects an abnormality, a push notification is sent to the user's smartphone. If the user checks the notification and identifies that the transaction was not initiated by them, the emotion engine detects that the user is extremely anxious and provides gentle instructions on how to respond. The user can also press a button in the app to immediately report the incident to their credit card company and take steps to suspend their card.

[1486] Continuous model updates

[1487] The server periodically updates transaction data and shipping destination information, and retrains the generative AI model to maintain or improve its accuracy. This allows the system to always perform highly accurate fraud detection based on the latest data. Furthermore, emotion data from an emotion engine is added, enabling more personalized responses.

[1488] Examples:

[1489] The server learns new fraud patterns through monthly data updates and retrains the model, ensuring the system can respond to the latest fraud techniques. Data from the emotion engine is also incorporated into the model, making it possible to customize notification methods and countermeasures based on the user's emotional state.

[1490] In this way, the system of the present invention detects fraudulent credit card use and fraudulent changes to shipping addresses in real time with high accuracy, and provides prompt notification and response that takes into account the user's emotional state, thereby realizing a safe and secure transaction environment for users and reducing their mental burden.

[1491] The processing flow will be explained below.

[1492] Step 1:

[1493] The server periodically collects credit card transaction data and shipping information.

[1494] Specific operation: Using the APIs of credit card companies and delivery companies, the system obtains the latest transaction data and delivery address information, while also collecting data on known past fraudulent use.

[1495] Step 2:

[1496] The server normalizes the collected data, removes outliers, and extracts features.

[1497] Specific operations: Extract key fields such as date, time, amount, store used, and delivery destination from raw data, normalize them into a consistent format, detect and remove outliers, and calculate the features required for analysis.

[1498] Step 3:

[1499] The server trains a generative artificial intelligence model using the preprocessed data.

[1500] Specific operation: The normalized and feature-extracted data is split into a training set and a test set. A generative AI model is trained to learn normal usage patterns and the characteristics of fraudulent usage.

[1501] Step 4:

[1502] When a user makes a transaction using a credit card, the terminal transmits the transaction data to the server in real time.

[1503] Specific operation: The POS system or online payment platform immediately sends the user's transaction data (e.g., time, location, amount, etc.) to the server.

[1504] Step 5:

[1505] The server immediately analyzes the received transaction data using a generative artificial intelligence model and detects any anomalies.

[1506] How it works: Incoming data is fed into a generative AI model in real time, where it is compared with normal usage patterns. If an abnormal pattern is detected, an anomaly flag is raised.

[1507] Step 6:

[1508] If an abnormality is detected, the server will promptly notify the user.

[1509] Specific behavior: If an abnormality flag is raised, a warning message will be sent using the user's contact information (email address, SMS, push notification, etc.).

[1510] Step 7:

[1511] The emotion engine analyzes the user's reaction in real time when they receive a notification and detects their emotions (e.g., surprise, anxiety, anger, etc.).

[1512] Specific operation: The emotion engine analyzes the user's voice, facial expressions, text input, etc. when receiving the notification to determine their emotional state.

[1513] Step 8:

[1514] The user receives a notification and verifies the authenticity of the transaction.

[1515] Specific operation: Upon receiving the notification, the user accesses the web portal or smartphone app to verify whether the transaction was initiated by them.

[1516] Step 9:

[1517] The server customizes the notification content and response procedures based on the user's emotions detected by the emotion engine.

[1518] Specific behavior: If the user's emotions indicate anxiety or surprise, the server will re-notify them with softer language and reassuring messages, and if necessary, instruct them to escalate the issue to a support staff member.

[1519] Step 10:

[1520] If the user determines that the use has been fraudulent, procedures such as suspending the card or changing the delivery address will be implemented.

[1521] Specific operation: If a user determines that a transaction has been fraudulent, they can contact their credit card company using a dedicated app or a link in the web portal to suspend their card. They can also contact the delivery company to suspend or change the delivery of their product.

[1522] Step 11:

[1523] The server periodically updates transaction data and delivery destination information and retrains the generative artificial intelligence model.

[1524] How it works: Regularly prepare new datasets and evaluate the performance of existing generative AI models. Retrain as needed to maintain or improve model accuracy. Additionally, by adding data from the emotion engine, the model can be trained to provide more personalized responses.

[1525] Through these steps, the system of the present invention detects credit card fraud and shipping fraud in real time and enables quick and appropriate responses based on the user's emotional state.

[1526] Example 2

[1527] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1528] In modern society, credit card fraud is on the rise, with the risk increasing especially in online shopping and remote transactions. When fraud occurs, it not only causes financial loss to users, but also causes significant psychological stress. Furthermore, existing fraud detection systems are often inadequate because they struggle to respond in real time and are unable to take into account the user's emotional state. This makes it difficult for many users to conduct transactions with confidence. Therefore, there is a need for a system that can detect fraud quickly and accurately, and provide notifications and responses that take the user's emotions into account.

[1529] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for collecting credit card transaction data and delivery destination information, means for normalizing the transaction data and delivery destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for collecting and analyzing user emotion data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, and means for notifying the user of detected anomalies and allowing the user to confirm the anomaly. This makes it possible to quickly and accurately detect fraudulent credit card use and to notify and respond in consideration of the user's emotional state.

[1530] "Credit card transaction data" means information that records purchases and payments made using a credit card.

[1531] "Delivery destination information" refers to information such as the address and contact details of the recipient of the purchased product or service.

[1532] "Normalization" is the process of standardizing the range and units of data to make it easier to compare and analyze.

[1533] "Outlier removal" is the process of identifying and eliminating invalid data points within a data set.

[1534] "Features" are important input variables used in machine learning models for prediction and classification.

[1535] A "generative artificial intelligence model" is an AI model that has the ability to learn from data, recognize patterns, and generate or analyze new data.

[1536] "User emotion data" refers to information about the user's emotional state obtained from the user's facial expression, voice, text input, and the like.

[1537] "Receiving transaction data in real time" means that the data is sent to the server and received immediately the moment a transaction occurs.

[1538] "Anomaly detection" refers to identifying fraudulent or unusual behavior that deviates from normal patterns.

[1539] "Notification" is the act of the system informing the user of abnormalities or important information.

[1540] This invention relates to a highly accurate and rapid detection system for preventing fraudulent use of credit cards. This system collects credit card transaction data and shipping destination information, and uses this data to learn and apply a generative artificial intelligence model that detects fraudulent use. The details are as follows:

[1541] Data collection and preprocessing

[1542] The server collects credit card transaction data and shipping information for each transaction, which involves retrieving the data through an API and storing it in a database using MySQL, a popular relational database management system (RDBMS).

[1543] Specifically, when a user purchases an item from an online shop, the transaction data (purchase date and time, amount, store location, etc.) and shipping information are stored in a MySQL database. The collected data is read using the Python Pandas library, normalized, and outliers are removed, preparing the data for processing.

[1544] Model training and generation

[1545] The server uses the preprocessed data to train a generative artificial intelligence model. This model learns normal usage patterns and the characteristics of fraudulent usage, and detects anomalies with high accuracy. The specific software used is Scikit-learn's RandomForestClassifier.

[1546] The model training uses transaction data and fraudulent use data collected to date. Additionally, Affectiva's SDK is used as a sentiment analysis library to collect and analyze user sentiment data. This sentiment data is also fed back into the model training.

[1547] Specifically, the server analyzes past transaction data to learn, for example, which region a user typically spends money in. At the same time, it analyzes emotional data using Affectiva's SDK and incorporates that data into the model.

[1548] Real-time detection

[1549] When a user purchases a product, the device sends the transaction data to the server in real time. The server immediately analyzes the received data using a generative artificial intelligence model and immediately generates a warning if an abnormality is detected. The server also analyzes the user's reactions using an emotion engine to detect the user's level of stress and anxiety.

[1550] Specifically, if a user makes a sudden, expensive purchase in a specific major city while traveling abroad, the transaction data is sent to the server, and the model determines that the transaction is different from the usual pattern. If it determines that fraudulent use is suspected, emotion analysis can also detect the user's anxiety.

[1551] User notification and response

[1552] The server uses Firebase Cloud Messaging as a communication library to promptly notify the user if an abnormality is detected. The user who receives the notification can check whether the transaction is legitimate, and if it is fraudulent, can immediately take action such as suspending the card or changing the delivery address.

[1553] Specifically, when the server detects an abnormality, it sends a push notification to the user's smartphone. If the user checks the notification and determines that the transaction was not initiated by them, the app uses emotional analysis to detect that the user is extremely anxious and provides gentle guidance. The app also allows the user to press a button to report the transaction to their credit card company and immediately take steps to suspend their card.

[1554] Continuous model updates

[1555] The server periodically updates transaction data and delivery destination information, and performs re-training to improve the accuracy of the generative AI model. This ensures highly accurate fraud detection based on the latest data. Emotional data is also added, enabling more personalized responses.

[1556] Specifically, the server updates data monthly, learns new fraudulent usage patterns, and retrains the model. This ensures the system can respond to the latest fraud techniques. Emotional data is also fed back to the model, allowing it to provide notification methods and countermeasures tailored to the user.

[1557] Prompt Sentence Examples

[1558] Fraud detection is performed on newly collected transaction data, including:

[1559] Purchase date: 2023-10-01 10:00:00

[1560] Amount: $500

[1561] Shop location: London

[1562] Delivery Address: 123 ABC Street, London

[1563] User emotion data: Moderate anxiety

[1564] Use this data to detect anomalies in real time using fraud detection models and notify users as needed.

[1565] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1566] System program processing flow

[1567] Step 1:

[1568] The server collects credit card transaction data and shipping information. Specifically, it uses an API to retrieve credit card transaction information and stores it in a MySQL database. The input includes the transaction data and shipping information retrieved from the API. The output is this raw data stored in the database.

[1569] Specific behavior:

[1570] Make API calls to retrieve transaction data and shipping information.

[1571] The retrieved data is stored in a MySQL database.

[1572] Step 2:

[1573] The server pre-processes the collected data. It uses the Pandas library to normalize the data, remove outliers, and extract features. The input includes raw data stored in a MySQL database. The output is the pre-processed, feature-extracted data.

[1574] Specific behavior:

[1575] Use Pandas to read data from a MySQL database.

[1576] Normalize the data to make it consistent across ranges.

[1577] Detect and remove outliers.

[1578] Important features (e.g., purchase date and time, amount, store location) are extracted and saved as structured data.

[1579] Step 3:

[1580] The server uses the preprocessed data to train a generative AI model. It uses Scikit-learn's RandomForestClassifier to learn normal transaction patterns and fraudulent usage patterns. The input includes training data with extracted features. The output is a trained generative AI model.

[1581] Specific behavior:

[1582] Define a model using Scikit-learn's RandomForestClassifier.

[1583] The data from which the features have been extracted is supplied to the model as training data and fitted.

[1584] The model is evaluated and parameters are adjusted as necessary.

[1585] Step 4:

[1586] The server collects and analyzes the user's emotional data. Using Affectiva's SDK, it analyzes the user's emotions (e.g., stress, anxiety, surprise, etc.) and stores the data. Inputs include the user's facial expressions, voice, and text input. The output is the analyzed emotional data.

[1587] Specific behavior:

[1588] Affectiva's SDK is used to analyze the user's facial expressions, voice, and text input.

[1589] The analysis results (emotion data) are saved as structured data and provided to a generative artificial intelligence model.

[1590] Step 5:

[1591] When a user purchases a product, the device sends transaction data to a server in real time. The server analyzes the received data using a generative artificial intelligence model to detect anomalies. The input includes real-time transaction data and a pre-trained model. The output is a warning message or notification when an anomaly is detected.

[1592] Specific behavior:

[1593] The terminal collects transaction data and immediately transmits it to the server.

[1594] The server receives the transaction data and inputs it into a generative artificial intelligence model.

[1595] If the model detects an anomaly, it generates an anomaly alert.

[1596] Step 6:

[1597] The server notifies the user of the detected anomaly, and the user confirms the anomaly. The push notification is sent using Firebase Cloud Messaging. The input includes the transaction data and sentiment data for which the anomaly was detected. The output is the notification and confirmation for the user.

[1598] Specific behavior:

[1599] Use Firebase Cloud Messaging to send push notifications to users' devices.

[1600] Users can review notifications and approve or report unusual transactions within the app.

[1601] Step 7:

[1602] The server periodically updates the transaction data and shipping destination information and retrains the generative AI model to improve its accuracy. The input includes new transaction data and updated emotion data. The output is a retrained, highly accurate generative AI model.

[1603] Specific behavior:

[1604] Extract new data from a MySQL database.

[1605] Retrain the model and reevaluate its performance.

[1606] Adjust parameters as needed to improve the accuracy of the model.

[1607] (Application example 2)

[1608] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1609] Fraudulent credit card use is becoming more sophisticated every year, making it difficult to quickly and accurately detect fraudulent transactions using conventional methods. Furthermore, when a fraudulent transaction is detected, notifications are given to users uniformly, and no response is given based on the user's emotional state. This often leaves users feeling anxious and stressed. The present invention aims to solve these problems and realize effective fraud detection and smooth user response.

[1610] The identification processing by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting credit card transaction data and delivery destination information, means for normalizing the transaction data and delivery destination information, removing outliers, and extracting features, means for training a generative artificial intelligence model using the normalized and feature-extracted data, means for collecting and analyzing user emotion data, means for receiving transaction data in real time and detecting anomalies using the generative artificial intelligence model, and means for notifying the user of detected anomalies, customizing the content of the notification based on the user's emotional state, and allowing the user to confirm the anomaly. This enables highly accurate real-time detection of fraudulent transactions and enables appropriate notification and response tailored to the user's emotional state.

[1611] Output for definition statements

[1612] "Credit card transaction data" refers to information about transactions conducted using a credit card, including details such as the transaction date and time, transaction amount, and transaction store.

[1613] "Shipping destination information" refers to information about the shipping destination of the purchased product, specifically information such as address, name, and contact details.

[1614] "Normalization" refers to the process of maintaining data consistency and ensuring consistency across different data sets.

[1615] An "outlier" is a data point that falls outside the normal range, such as an extremely high or low value in a data set.

[1616] "Features" refer to specific attributes or parameters of input data that a generative artificial intelligence model uses for learning and prediction.

[1617] A "generative artificial intelligence model" is a statistical or machine learning model that learns patterns and features based on large amounts of data and makes predictions and classifications.

[1618] "Emotional data" refers to information about the user's emotional state obtained from their voice, facial expression, text input, etc.

[1619] "Real-time" refers to immediate processing or reaction, meaning data is collected, analyzed, and responded to with short latency.

[1620] "Anomaly detection" refers to the process of automatically identifying fraudulent activity or anomalous behavior that deviates from normal usage patterns.

[1621] "Customizing notification content" refers to changing the message and method of notification to a user to suit the status and characteristics of each individual user.

[1622] MODE FOR CARRYING OUT THE INVENTION

[1623] Overall system overview

[1624] The system of the present invention includes components for detecting fraudulent credit card use with high accuracy and for providing prompt and appropriate responses to users. The system is primarily composed of a server, terminals, and users.

[1625] Server Processing

[1626] The server processes the data in the following steps to detect and notify anomalies.

[1627] 1. Data Collection

[1628] The server collects credit card transaction data and shipping information, including details such as transaction date and time, amount, currency, merchant, and shipping address.

[1629] Hardware: The servers are general-purpose servers equipped with high-performance processors and storage.

[1630] Software: Data is collected through a database management system (DBMS) or API.

[1631] 2. Data Preprocessing

[1632] The collected data is normalized, outliers are removed, and features are extracted, providing data suitable for training generative AI models.

[1633] Software used: Data preprocessing scripts using programming languages ​​such as Python and R, and data manipulation libraries such as Pandas and NumPy.

[1634] 3. Learning generative AI models

[1635] The normalized and feature-extracted data is used to train a generative AI model, which learns the characteristics of normal and fraudulent usage patterns.

[1636] Hardware used: High-performance server equipped with GPU etc.

[1637] Software used: Machine learning libraries such as TensorFlow and PyTorch.

[1638] 4. Emotional Data Collection and Analysis

[1639] It collects and analyzes emotional data such as voice and facial expressions from users' smartphones and other devices, uses an emotion engine to identify emotional states, and then applies that data to generative AI models.

[1640] Hardware used: Smartphone camera and microphone.

[1641] Software used: Speech recognition API and image analysis API.

[1642] 5. Real-time detection and notification

[1643] It receives real-time data as transactions occur, analyzes it with a generative AI model, and sends customized notifications based on the user's emotional state if an anomaly is detected.

[1644] Hardware used: High-performance server capable of real-time processing.

[1645] Software used: Real-time data streaming framework, push notification service.

[1646] Specific examples

[1647] When a user makes a transaction using a credit card, the data (e.g., the transaction amount is 150,000 yen, and the transaction location is New York) is sent to the server. The server immediately analyzes this transaction data using a generative AI model and compares it with normal usage patterns. If the transaction is determined to be abnormal, the server uses an emotion engine to obtain emotional data from the smartphone's camera and microphone and detects that the user is anxious. As a result, it can send a notification adapted to the user's emotional state, such as displaying a message saying, "There is a suspicion of a fraudulent transaction. Don't worry, we will take immediate action."

[1648] Prompt Sentence Examples

[1649] Here are some examples of prompts for generative AI models:

[1650] Please determine whether the following credit card transaction is legitimate. Transaction information: Amount: 150,000 yen, Location: New York, User: Tokyo. Is this transaction different from the usual pattern? Also, the user's voice indicates anxiety.

[1651] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1652] Program processing steps

[1653] Step 1:

[1654] The server collects all credit card transaction data and shipping information. For each transaction, details such as transaction date and time, transaction amount, currency, transaction store, shipping address, etc. are stored in a database. This input data is recorded as a record for each transaction.

[1655] Step 2:

[1656] The server normalizes the collected transaction data and removes outliers. This includes maintaining data consistency and standardizing the format. For example, it detects and removes transactions with unusually high prices or inaccurate shipping information. This process produces clean, consistent feature data.

[1657] Step 3:

[1658] Using the normalized and feature-extracted data, the server trains a generative AI model. This model learns past patterns of normal and fraudulent usage. The data inputs are historical transaction data and its features, and the output is feedback that the model uses to classify normal and abnormal behavior.

[1659] Step 4:

[1660] A user's smartphone or other device collects and sends emotional data to a server. Emotional data includes information such as the user's voice and facial expressions. The emotion engine analyzes this data and identifies the user's emotional state (e.g., anxiety, stress). This data is the input data obtained through the emotion engine API, and the analyzed emotional state is the output.

[1661] Step 5:

[1662] The server receives transaction data in real time and analyzes it using a generative AI model. Here, detailed transaction information is the input data, and the output is an anomaly detected as the analysis result of the AI ​​model. If the AI ​​detects an anomaly, the details are immediately passed on to the next step.

[1663] Step 6:

[1664] The server notifies the user of detected anomalies and customizes the notification content based on the user's emotional state. The notification message is generated based on the user's emotional state obtained by the emotion engine and the transaction data. For example, if the server detects that the user is feeling anxious or stressed, the notification content is customized with gentler language such as "Don't worry, we will take immediate action." The output data of the emotion engine and the abnormal transaction data are used as input, and the output is a notification message to the user.

[1665] Step 7:

[1666] The user receives a notification through their device and checks whether the abnormal transaction is legitimate. If the user checks the transaction and determines it to be fraudulent, they can send instructions to the server via the app, such as suspending their credit card or changing the shipping address. This inputs the user's confirmation data, and the terminal then sends instructions to the card company, etc., which is the output.

[1667] Step 8:

[1668] The server periodically updates transaction data and delivery destination information, retraining the generative AI model to maintain or improve its accuracy. Emotional data is also periodically added and incorporated into the model. This allows the system to always use the latest information for highly accurate fraud detection. The latest transaction data and emotional data are used as input, and the output is an updated AI model.

[1669] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1670] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1671] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1672] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1673] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1674] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1675] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1676] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1677] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1678] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1679] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1680] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1681] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1682] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1683] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1684] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1685] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1686] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1687] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1688] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1689] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1690] The following is further disclosed regarding the above embodiment.

[1691] (Claim 1)

[1692] a means for collecting credit card transaction data and shipping information;

[1693] means for normalizing the transaction data and delivery destination information, removing outliers, and extracting features;

[1694] A means for training a generative artificial intelligence model using the normalized and feature-extracted data;

[1695] a means for receiving transaction data in real time and detecting anomalies using a generative artificial intelligence model;

[1696] a means for notifying a user of the detected abnormality and allowing the user to confirm the abnormality;

[1697] A system including:

[1698] (Claim 2)

[1699] 10. The system of claim 1, wherein the generative artificial intelligence model is adapted to learn normal usage patterns and characteristics of fraudulent usage.

[1700] (Claim 3)

[1701] 10. The system of claim 1, further comprising means for periodically updating the transaction data and shipping information, assessing the accuracy of the generative artificial intelligence model, and retraining it as necessary.

[1702] "Example 1"

[1703] (Claim 1)

[1704] a means for collecting credit card transaction data and payee information;

[1705] means for normalizing the transaction data and recipient information, removing outliers, and extracting features;

[1706] A means for training a generative artificial intelligence model using the normalized and feature-extracted data;

[1707] A means for receiving transaction data in real time and detecting anomalies using a generative artificial intelligence model;

[1708] a means for notifying the user of the detected abnormality, allowing the user to confirm the abnormality and, if fraudulent, suspending the payment method or changing the recipient information;

[1709] means for periodically updating the transaction data and recipient information and re-learning the generative artificial intelligence model to maintain or improve its accuracy;

[1710] A system including:

[1711] (Claim 2)

[1712] 10. The system of claim 1, wherein the generative artificial intelligence model is adapted to learn normal usage patterns and characteristics of fraudulent usage.

[1713] (Claim 3)

[1714] 10. The system of claim 1, further comprising means for periodically updating the transaction data and recipient information, assessing the accuracy of the generative artificial intelligence model, and retraining it as necessary.

[1715] "Application Example 1"

[1716] (Claim 1)

[1717] a means for collecting credit card transaction data and shipping information;

[1718] means for normalizing the transaction data and delivery destination information, removing outliers, and extracting feature values;

[1719] A means for training a generative artificial intelligence model using the normalized and feature-extracted data;

[1720] A means for receiving transaction data in real time and detecting anomalies using a generative artificial intelligence model;

[1721] means for sending a push notification of the detected abnormality to a user's communication terminal;

[1722] A means for users to immediately take steps to suspend their credit cards based on the results of the generative artificial intelligence model; and

[1723] A system including:

[1724] (Claim 2)

[1725] 10. The system of claim 1, wherein the generative artificial intelligence model is adapted to learn normal usage patterns and characteristics of fraudulent usage.

[1726] (Claim 3)

[1727] 10. The system of claim 1, further comprising means for periodically updating the transaction data and shipping information, assessing the accuracy of the generative artificial intelligence model, and retraining it as necessary.

[1728] "Example 2: Combining Emotion Engines"

[1729] (Claim 1)

[1730] a means for collecting credit card transaction data and shipping information;

[1731] means for normalizing the transaction data and delivery destination information, removing outliers, and extracting feature values;

[1732] A means for training a generative artificial intelligence model using the normalized and feature-extracted data;

[1733] means for collecting and analyzing user emotion data;

[1734] A means for receiving transaction data in real time and detecting anomalies using a generative artificial intelligence model;

[1735] a means for notifying a user of the detected abnormality and allowing the user to confirm the abnormalit...

Claims

1. a means for collecting credit card transaction data and shipping information; means for normalizing the transaction data and delivery destination information, removing outliers, and extracting features; A means for training a generative artificial intelligence model using the normalized and feature-extracted data; a means for receiving transaction data in real time and detecting anomalies using a generative artificial intelligence model; a means for notifying a user of the detected abnormality and allowing the user to confirm the abnormality; A system including:

2. 10. The system of claim 1, wherein the generative artificial intelligence model is adapted to learn normal usage patterns and characteristics of fraudulent usage.

3. 10. The system of claim 1, further comprising means for periodically updating the transaction data and shipping information, assessing the accuracy of the generative artificial intelligence model, and retraining it as necessary.

Images