System

Abstract

A system is provided.SOLUTION: A system including means for collecting risk-related information from a data source, means for learning and constructing a risk detection model based on the collected information, means for detecting a risk in real time using the learned risk detection model, means for generating a risk alert based on the detected risk and notifying a person in charge, means for displaying the generated risk alert and a recommended countermeasure on a user terminal, and means for feeding back an execution result of the risk countermeasure based on the risk alert.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] Companies and organizations are required to comply with complex laws and regulations and manage risks, which requires a huge amount of time and manpower. As risk management becomes increasingly difficult due to the expansion of business domains and the increase in new businesses, current systems make it difficult to identify new risks and respond quickly. As a result, violations of laws and regulations or unethical behavior may be overlooked, increasing the risk to a company's reputation. To solve this problem, an efficient and automated risk management system is required. [Means for solving the problem]

[0005] The present invention solves the above-mentioned problems by providing a system that includes: means for collecting risk-related information from data sources; means for training and building a risk detection model based on the collected information; means for detecting risks in real time using the trained risk detection model; means for generating risk alerts based on the detected risks and notifying personnel; means for displaying the generated risk alerts and recommended countermeasures on a user terminal; and means for providing feedback on the results of the risk countermeasures implemented based on the risk alerts. Furthermore, the system includes means for preprocessing the collected data and converting it into a format suitable for training an AI model, means for training the AI ​​model based on the preprocessed data, and means for evaluating the accuracy of the trained AI model and retraining it as necessary, thereby maintaining and improving the accuracy of the model. This improves the accuracy of risk detection and enables efficient resolution of complex risk management problems faced by companies and organizations.

[0006] "Data Sources" refers to internal systems and external information sources that provide risk-related information.

[0007] "Collect" refers to the act of obtaining necessary risk-related information from data sources.

[0008] "Risk detection model" refers to the machine learning or AI-based algorithm used to analyze collected data and automatically identify risks.

[0009] "Learning" refers to the training process to improve the performance of a risk detection model based on acquired data.

[0010] "Real-time" means that data is processed and analyzed the moment it is generated, providing immediate results.

[0011] "Risk detection" refers to the process by which potential risks are identified by a risk detection model.

[0012] "Risk Alert" refers to a warning message sent to notify personnel about a detected risk.

[0013] "Notification" refers to the means by which generated risk alerts are communicated to responsible parties.

[0014] "User Terminal" means a computing device used to display risk alerts and other information.

[0015] "Feedback" refers to the process of returning the results of risk response measures to the system, providing information that will help improve the performance of AI models.

[0016] "Preprocessing" refers to operations such as data cleansing and normalization that transform raw data into a format suitable for training an AI model.

[0017] "Parameter tuning" refers to the optimization of hyperparameters in training an AI model.

[0018] A "risk score" refers to a numerical representation of the impact or importance of a detected risk.

[0019] "Recommended measures" refers to specific actions to be taken in response to detected risks.

[0020] "Retraining" refers to additional training of an existing model by adding new data or improvements to improve its performance. [Brief explanation of the drawings]

[0021] [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

[0022] 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.

[0023] First, the terms used in the following description will be explained.

[0024] 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).

[0025] 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.

[0026] 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.

[0027] 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.

[0028] 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."

[0029] [First embodiment]

[0030] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0031] 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.

[0032] 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).

[0033] 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.

[0034] 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.

[0035] 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.

[0036] 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.

[0037] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0038] 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.

[0039] 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.

[0040] 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.

[0041] 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."

[0042] The present invention relates to a system for enabling a company to efficiently manage risk. The specific processing content of the program of this system and an embodiment thereof will be described below with specific examples.

[0043] Overall flow

[0044] This system operates between the server, terminals, and users as follows: The server collects and analyzes data, trains AI models, and detects risks and generates alerts. The terminals are responsible for displaying alert notifications and risk information, while users implement risk countermeasures and provide feedback.

[0045] 1. Data Collection and Analysis

[0046] Server Action:

[0047] 1. Data collection: The server collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external legal databases and social networking sites.

[0048] 2. Data Update: The server saves and updates the collected data to the database used by date and time.

[0049] 2. Learning and training the AI ​​model

[0050] Server Action:

[0051] 1. Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the AI ​​model.

[0052] 2. Model training: The server uses the pre-processed data to train an AI model for risk detection, leveraging machine learning algorithms (e.g., random forest, deep learning).

[0053] 3. Model evaluation: The server evaluates the accuracy of the model and retrains it if necessary.

[0054] 3. Risk detection and alert generation

[0055] Server Action:

[0056] 1. Real-time data analysis: The server analyzes incoming business data in real time and detects risks using AI models.

[0057] 2. Risk Scoring: The server calculates a risk score for each detected risk and evaluates its severity.

[0058] 3. Alert generation: If the risk score exceeds a certain threshold, the server generates an alert and notifies the appropriate personnel.

[0059] 4. Alert notification and display

[0060] Terminal handling:

[0061] 1. Receiving an alert: The terminal receives an alert sent from the server.

[0062] 2. Alert display: The terminal displays detailed information about the alert (risk details, scope of impact, recommended countermeasures) to the user.

[0063] 5. Risk Management

[0064] User Action:

[0065] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[0066] 2. Implementing countermeasures: The user implements the necessary risk countermeasures based on the recommended countermeasures presented, such as suspending transactions, reviewing processes, or initiating an internal investigation.

[0067] 3. Providing feedback: The user provides feedback on the results of the implementation to the server, and this information is used to retrain the model.

[0068] Specific examples

[0069] As an example, consider a scenario in which Company A detects reputation risks, including negative reactions on social media, when releasing a new product.

[0070] 1. Data Collection: The server collects social media data and internal marketing data.

[0071] 2. Training the AI ​​model: Train the risk detection model based on the pre-processed data.

[0072] 3. Risk detection: When a new product is released, the server detects negative feedback in real time and calculates a high risk score.

[0073] 4. Alert Generation: The server generates and notifies the marketing department personnel of a risk alert.

[0074] 5. Notification and response: The device receives an alert, the user checks the relevant information, and modifies the new marketing strategy based on the recommendations.

[0075] In this way, the present invention enables efficient management of the complex risks faced by companies and realizes prompt and appropriate responses.

[0076] The processing flow will be explained below.

[0077] Step 1: Data collection and updates

[0078] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external data sources (e.g., legal databases, SNS). The collected data is stored and updated in a database.

[0079] Step 2: Data Preprocessing

[0080] The server cleans the collected data, fills in missing data, removes outliers, and then standardizes (normalizes) the data and converts it into a format suitable for training AI models.

[0081] Step 3: Training the AI ​​model

[0082] The server uses the preprocessed data to train the AI ​​model, applying machine learning algorithms to learn patterns to identify risks. After training, the model's accuracy is evaluated and retrained if necessary.

[0083] Step 4: Real-time data analysis

[0084] The server uses AI models to detect risks as business data flows in in real time, using data stream analysis techniques to identify outliers and unexpected patterns.

[0085] Step 5: Risk Scoring

[0086] The server calculates a risk score for each detected risk, based on the scope of impact and probability of occurrence, and serves as a basis for assessing the severity of the risk.

[0087] Step 6: Alert Generation

[0088] The server generates risk alerts when the risk score exceeds a pre-defined threshold, with detailed information about the risk, its scope, and recommended actions to take.

[0089] Step 7: Alert Notifications

[0090] The terminal receives risk alerts sent from the server, and when an alert arrives, it notifies the user so that they can check it immediately.

[0091] Step 8: Displaying alerts

[0092] The device displays detailed information about the risk alert to the user, including the type of risk, the scope of impact, and specific recommended actions to take.

[0093] Step 9: Implement risk responses

[0094] Based on the displayed risk alerts and recommended actions, users can take necessary risk response measures, such as suspending trading, reviewing processes, or launching an internal investigation.

[0095] Step 10: Provide feedback

[0096] Users then provide feedback on the results of their risk response measures to the server, which uses the feedback information to subsequently retrain the AI ​​model, contributing to improving the model's accuracy.

[0097] Example 1

[0098] 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."

[0099] Companies are being asked to improve the efficiency of risk management and respond quickly, but many challenges exist in collecting information from diverse data sources, training AI models, detecting and notifying risks in real time, and incorporating feedback. In particular, the processes of data preprocessing and model retraining are complex, and integrating the entire system takes time and effort. Furthermore, real-time risk detection and accurate alert generation are difficult, which can lead to a lack of speed and accuracy in risk responses. It is desirable to resolve these issues.

[0100] 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.

[0101] In this invention, the server includes a means for collecting risk-related information from data sources, a means for storing and updating the collected information in a database by date and time, and a means for preprocessing the collected data and converting it into a format suitable for training a generative AI model, thereby enabling companies to efficiently manage risk using an integrated system.

[0102] "Data Source" is a collective term for internal systems and external sources used to collect risk-related information.

[0103] A "risk detection model" is a general term for algorithms and machine learning models that are learned and built based on collected data to detect risks.

[0104] "Real-time" refers to processing and analysis being carried out immediately, close to the moment data is generated.

[0105] "Risk alert" refers to a warning or caution message that is generated based on a detected risk and notified to a responsible person.

[0106] "User Terminal" refers collectively to any electronic device used to display risk alerts and recommended responses.

[0107] "Feedback" refers to providing information about the results of risk response measures to the server.

[0108] A "generative AI model" refers to a model built using machine learning algorithms based on collected data.

[0109] "Preprocessing" is the process of organizing and transforming collected data to convert it into a form suitable for training an AI model.

[0110] An "algorithm" refers to a combination of steps or computational methods for solving a specific problem.

[0111] A "hyperparameter" is a setting value during the training process of a machine learning model, and refers to an element that affects the performance of the model.

[0112] A "stream processing framework" refers to a software platform for processing and analyzing data in real time.

[0113] The present invention relates to a system for enabling a company to efficiently manage risk. The specific processing contents of the program of this system and an embodiment thereof will be described below.

[0114] Overall flow

[0115] This system operates between the server, terminals, and users as follows: The server collects and analyzes data, trains generative AI models, and detects risks and generates alerts. The terminals are responsible for notifying alerts and displaying risk information, while users implement risk countermeasures and provide feedback.

[0116] Data collection and analysis

[0117] Server Processing

[0118] 1. Data collection: The server obtains business data from internal systems (e.g., ERP systems, CRM systems) and also collects risk-related information from external legal databases and social media APIs.

[0119] The server obtains financial and supply chain data by calling the ERP system's API.

[0120] Use a social media API (e.g., Twitter API) to collect tweets related to a specific hashtag.

[0121] Use a web scraping tool (e.g., BeautifulSoup) to scrape new legal information from the legal database.

[0122] 2. Data update: The server saves and updates the collected data in a database (e.g., MySQL (registered trademark), PostgreSQL) according to the date and time.

[0123] The new data is stored in a temporary table and then reconciled with the existing database using a journal file to ensure data consistency.

[0124] If the data is updated, update the timestamp and insert the new data.

[0125] Learning and training AI models

[0126] Server Processing

[0127] 1. Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the generative AI model, specifically by imputing missing data, converting character data to numeric data, and normalizing the data.

[0128] Missing data will be imputed using the median.

[0129] Character data is converted to numeric data using one-hot encoding.

[0130] Normalization is performed to align the range of the data.

[0131] 2. Model training: The preprocessed data is used to train a generative AI model. For model training, a machine learning framework such as Tensorflow (registered trademark) or scikit-learn is used, using, for example, random forest or deep learning algorithms.

[0132] The dataset is divided into batches for training.

[0133] Tune the hyperparameters of your model to get optimal performance.

[0134] 3. Model evaluation: Evaluate the accuracy of the trained model and retrain it if necessary.

[0135] For evaluation, we use a test dataset and calculate precision, recall, and F1 score.

[0136] If the accuracy is low, reset the hyperparameters and retrain.

[0137] Risk detection and alert generation

[0138] Server Processing

[0139] 1. Real-time data analysis: The server analyzes incoming business data in real time and detects risks using generative AI models, utilizing stream processing frameworks such as Apache Kafka and Apache Flink.

[0140] Monitor real-time data streams and detect specific patterns.

[0141] Input data into the AI ​​model to obtain a predicted value.

[0142] 2. Risk Scoring: The server calculates a risk score for each detected risk and evaluates its severity.

[0143] Set scoring criteria according to the type of risk.

[0144] A score is calculated and evaluated based on the output value of the AI ​​model.

[0145] 3. Alert generation: If the risk score exceeds the set threshold, the server generates an alert and notifies the responsible person.

[0146] An email containing the alert content is created and sent using the SAP communication protocol.

[0147] Use a push notification service to send alerts to the mobile devices of personnel.

[0148] Alert Notification and Display

[0149] Terminal handling

[0150] 1. Receiving an alert: The device receives an alert sent from the server.

[0151] Receives push notifications from the notification server and displays them on the device.

[0152] 2. Alert display: Displays detailed information about the received alert (risk content, scope of impact, recommended countermeasures) to the user.

[0153] The mobile or desktop app will pop up an alert message.

[0154] Risk Management

[0155] User Action

[0156] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[0157] Check the alert displayed on the device screen in detail.

[0158] 2. Implementing Actions: The user implements the necessary risk response measures based on the recommended actions, which may include suspending trading, reviewing processes, or initiating an internal investigation.

[0159] Implement countermeasures based on company policy.

[0160] 3. Providing feedback: The results of the implemented countermeasures are provided to the server as feedback.

[0161] Feedback data is sent to the server using a dedicated input form.

[0162] Specific examples

[0163] Consider a scenario in which Company A detects reputation risks, including negative reactions on social media, when releasing a new product.

[0164] 1. Data Collection: The server collects social media data and internal marketing data.

[0165] 2. Training a generative AI model: Train a risk detection model based on the preprocessed data.

[0166] 3. Risk detection: When a new product is released, the server detects negative feedback in real time and calculates a high risk score.

[0167] 4. Alert generation: The server generates and notifies the marketing department personnel of a risk alert.

[0168] 5. Notification and response: The device receives an alert, the user checks the relevant information, and then a new marketing strategy is modified based on the recommendations.

[0169] Example prompt: "Describe a system that detects negative reactions on social media in real time when a company releases a new product and notifies the marketing department."

[0170] In this way, the present invention enables efficient management of the complex risks faced by companies and realizes prompt and appropriate responses.

[0171] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0172] Step 1: Data collection

[0173] The server retrieves business data from internal systems (e.g., ERP systems, CRM systems) and also collects risk-related information using external legal databases and SNS APIs.

[0174] Input: Business data, legal data, and social media data from data sources.

[0175] Specific behavior:

[0176] The server obtains financial and supply chain data by calling the ERP system's API.

[0177] Use a social media API (e.g., Twitter API) to collect tweets related to a specific hashtag.

[0178] Use a web scraping tool (e.g., BeautifulSoup) to scrape new legal information from the legal database.

[0179] Output: The retrieved data.

[0180] Step 2: Data Update

[0181] The server stores and updates the collected data in a database (e.g., MySQL, PostgreSQL) according to the date and time.

[0182] Input: The retrieved data.

[0183] Specific behavior:

[0184] Save the new data in a temporary table and compare it with the existing database to ensure data consistency during the update.

[0185] When data is updated, the timestamp is also updated and the latest data is saved.

[0186] Output: The updated database.

[0187] Step 3: Data Preprocessing

[0188] The server preprocesses the collected data and converts it into a format suitable for training the generative AI model, specifically by imputing missing data, converting character data to numeric data, and normalizing the data.

[0189] Input: Data from the updated database.

[0190] Specific behavior:

[0191] Missing data will be imputed using the median.

[0192] Character data is converted to numeric data using one-hot encoding.

[0193] Normalize the data to the range 0 to 1.

[0194] Output: Preprocessed data.

[0195] Step 4: Model training

[0196] The preprocessed data is used to train generative AI models, using machine learning frameworks including TensorFlow and scikit-learn.

[0197] Input: Preprocessed data.

[0198] Specific behavior:

[0199] Define a dataset and train batch by batch.

[0200] Set and tune the hyperparameters of your model for optimal performance.

[0201] Deep learning algorithms are used to train the models.

[0202] Output: A trained AI model.

[0203] Step 5: Model evaluation

[0204] Evaluate the accuracy of trained AI models and retrain them if necessary.

[0205] Input: A trained AI model, a test dataset.

[0206] Specific behavior:

[0207] The model is evaluated using a test dataset.

[0208] Calculate precision, recall, and F1 score to measure the performance of the model.

[0209] Adjust hyperparameters and retrain as needed.

[0210] Output: The evaluated AI model.

[0211] Step 6: Real-time data analysis

[0212] The server uses stream processing frameworks such as Apache Kafka and Apache Flink to analyze incoming business data in real time and detect risks using generative AI models.

[0213] Input: Business data flowing in in real time.

[0214] Specific behavior:

[0215] Monitor real-time data streams and detect important patterns.

[0216] The data is analyzed using an AI model to obtain predicted values.

[0217] Output: Risk detection results.

[0218] Step 7: Risk Scoring

[0219] The server calculates a risk score for each detected risk and evaluates its severity.

[0220] Input: Risk findings.

[0221] Specific behavior:

[0222] Use pre-defined scoring criteria for each type of risk.

[0223] A risk score is calculated based on the output value of the AI ​​model.

[0224] Output: Risk scoring results.

[0225] Step 8: Alert Generation

[0226] If the risk score exceeds a set threshold, the server generates an alert and notifies the responsible person.

[0227] Input: Risk scoring results.

[0228] Specific behavior:

[0229] Create and send an email containing the alert content.

[0230] Send alerts to personnel's mobile devices via push notification services.

[0231] Output: The generated alert.

[0232] Step 9: Receiving and viewing alerts

[0233] The terminal receives the alert sent from the server and displays the detailed information to the user.

[0234] Input: The generated alert.

[0235] Specific behavior:

[0236] Receives push notifications from the notification server and displays them on the device.

[0237] Pop up an alert message in the mobile or desktop app.

[0238] Output: The displayed alert.

[0239] Step 10: Risk response

[0240] The user checks the contents of the alert displayed on the terminal and implements the necessary risk countermeasures based on the recommended countermeasures.

[0241] Input: The displayed alert.

[0242] Specific behavior:

[0243] Check the alert displayed on the device screen.

[0244] In accordance with our internal policies, we will take appropriate action, which may include suspending business, reviewing processes, or initiating an internal investigation.

[0245] Output: The results of the implemented measures.

[0246] Step 11: Provide feedback

[0247] The user provides the results of the risk countermeasures implemented to the server as feedback.

[0248] Input: The result of the action taken.

[0249] Specific behavior:

[0250] Feedback data is sent to the server using a dedicated input form.

[0251] The server stores the feedback data and uses it to improve the model in the future.

[0252] Output: The newly collected feedback data.

[0253] (Application example 1)

[0254] 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."

[0255] With the evolution of autonomous vehicles, it has become increasingly important to adapt to real-time changes in traffic conditions and the surrounding environment and quickly and appropriately detect and manage risks. However, with conventional technology, it has been difficult to efficiently analyze the large amounts of data obtained from sensors and cameras and appropriately evaluate and notify risks. This has resulted in delayed responses to risks such as traffic accidents and breakdowns, which could result in serious damage. Therefore, improvements to real-time risk management systems are required.

[0256] 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.

[0257] In this invention, the server includes means for collecting risk-related information from a data source, means for learning and building a risk detection model based on the collected information, means for detecting risks in real time using the learned risk detection model, means for generating a risk alert based on the detected risk and notifying a responsible person, means for displaying the generated risk alert and recommended countermeasures on a user terminal, means for feeding back the results of executing the risk countermeasures based on the risk alert, means for collecting data in real time from sensors and cameras of the autonomous vehicle and analyzing traffic conditions and the surrounding environment, and means for displaying and executing countermeasures in the autonomous vehicle based on the risk alert, thereby enabling real-time risk management of the autonomous vehicle.

[0258] "Data sources" are sources of information used to collect risk-related information, such as internal systems, external legal databases, and social media.

[0259] A "risk detection model" is a trained AI model for detecting risks based on collected data, using machine learning algorithms such as random forest and deep learning.

[0260] A "risk alert" is a warning message that is generated based on a detected risk and is sent to a responsible person.

[0261] "User terminal" refers to a device that displays alert notifications and risk information. Specifically, this applies to smartphones and head-mounted displays.

[0262] A "sensor" is a device installed in an autonomous vehicle that measures speed, distance, acceleration, etc.

[0263] The "camera" is an image capture device that visualizes the surrounding environment of an autonomous vehicle and is used to grasp the situation in real time.

[0264] "Traffic conditions" refers to information that comprehensively indicates the position, speed, direction, etc. of vehicles and pedestrians on the road.

[0265] "Surrounding environment" refers to the physical and dynamic elements surrounding an autonomous vehicle, including buildings, other vehicles, and pedestrians.

[0266] "Feedback" is information used to evaluate the results of risk response measures taken and to help with the next model training.

[0267] "Real-time" refers to the near-instantaneous collection, analysis, notification, and response of information.

[0268] "Recommended countermeasures" refers to specific countermeasures or action plans suggested to users in response to detected risks.

[0269] The present invention relates to a system that collects and analyzes data in real time from sensors and cameras in autonomous vehicles, and quickly and appropriately detects and manages risks according to traffic conditions and the surrounding environment. Specific embodiments for carrying out the present invention are described below.

[0270] This system mainly consists of three components: a server, a terminal, and a user.

[0271] server

[0272] The server is responsible for data collection, analysis, training of AI models, risk detection, and alert generation. Specifically, it has the following functions:

[0273] Data collection: The server collects data in real time from the autonomous vehicle's sensors and cameras. The hardware used here includes the autonomous vehicle's speed sensor, distance sensor, acceleration sensor, camera, etc.

[0274] Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the AI ​​model. Preprocessing includes data normalization and feature engineering. Software used includes Python, Pandas, and Numpy.

[0275] Training the AI ​​model: The server uses the preprocessed data to train a risk detection model. Machine learning algorithms include random forests and deep learning. Software used here includes scikit-learn.

[0276] Risk detection: Analyzes and detects risks in real time using trained AI models, and if a risk is detected, calculates a risk score and assesses its severity.

[0277] Alert generation: When a risk exceeds a certain threshold in importance, a risk alert is generated and the responsible person is notified.

[0278] Terminal

[0279] The device receives risk alerts from the server and notifies the user. Specifically, it has the following functions:

[0280] Receive alerts: Receive risk alerts sent from the server in real time.

[0281] Alert display: Received alerts are displayed visually and intuitively to the user on devices such as smartphones and head-mounted displays. These devices display detailed information about the alert (risk details, scope of impact, recommended countermeasures).

[0282] User

[0283] Users receive risk alerts via their devices and take appropriate countermeasures, which are detailed below:

[0284] Alert confirmation: The user confirms the content of the risk alert displayed on the device.

[0285] Countermeasure Execution: Based on the proposed countermeasures, implement the necessary risk countermeasures, such as applying emergency braking or proposing a route change.

[0286] Providing feedback: The results of the implemented countermeasures are fed back to the server. This information is used for the next retraining of the AI ​​model.

[0287] Examples and prompts

[0288] Examples:

[0289] When an autonomous vehicle is traveling on a highway, sensors detect a sudden decrease in the distance to the vehicle ahead. This data is pre-processed, and an AI model evaluates the risk in real time. If the risk is deemed high, an alert is generated. The alert is immediately displayed on a smartphone or head-mounted display, prompting the driver to apply emergency braking.

[0290] Prompt statement:

[0291] Write code to assess risk and generate an alert if the risk is high, based on the following autonomous vehicle sensor data:

[0292] Speed: float

[0293] Proximity: float

[0294] Acceleration: float

[0295] Include data pre-processing, training advanced AI models, data analysis, and alert generation as needed.

[0296] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0297] Step 1:

[0298] Data collection

[0299] The server collects real-time data from the autonomous vehicle's sensors and cameras, including speed, proximity, acceleration, etc. The collected data serves as input for subsequent processing steps.

[0300] Step 2:

[0301] Data Preprocessing

[0302] The server preprocesses the data collected in step 1. This preprocessing involves normalizing the data and performing feature engineering. For example, it normalizes the speed data to fit within a specified range and removes outliers. This preprocessed data serves as input for training and evaluating the AI ​​model.

[0303] Step 3:

[0304] Training an AI model

[0305] The server trains an AI model based on the preprocessed data from step 2. It uses algorithms such as random forest and deep learning to build a risk detection model. This trained model generates output results for detecting risks in real time.

[0306] Step 4:

[0307] Real-time risk detection

[0308] The server uses a trained AI model to analyze the data collected in real time and detect risks. For example, a sudden decrease in the distance ahead of the vehicle is considered a high risk. The results of this analysis serve as input data for subsequent alert generation.

[0309] Step 5:

[0310] Risk Scoring and Alerting

[0311] The server calculates a risk score and evaluates the severity based on the risks detected in step 4. If the risk score exceeds a certain threshold, it generates a risk alert and notifies the responsible party. The alert includes details of the risk, the scope of impact, and recommended countermeasures.

[0312] Step 6:

[0313] Alert Notification and Display

[0314] The device receives risk alerts sent from the server and notifies the user. The alerts are displayed via a user interface such as a smartphone or head-mounted display. Recommendations for risk countermeasures are also displayed on the device in real time.

[0315] Step 7:

[0316] Implementing risk response measures

[0317] The user checks the risk alerts and recommended countermeasures displayed on the device and takes specific countermeasures against the risk, such as applying emergency braking or proposing a route change. The results of this action become input data for the next feedback step.

[0318] Step 8:

[0319] Providing Feedback

[0320] Users can provide feedback on the results of their risk countermeasures to the server, which will be used to retrain the AI ​​model and improve the accuracy of the risk detection model.

[0321] 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.

[0322] The present invention relates to a system for companies to perform efficient and sophisticated risk management. In particular, by combining it with an emotion engine that recognizes user emotions, it provides a more accurate risk detection and alert system.

[0323] Overall flow

[0324] This system operates as follows between the server, terminal, and user: The server collects and analyzes data, trains AI models, detects risks, and generates alerts. The terminal is responsible for notifying alerts and displaying risk information, and recognizes user emotions using an emotion engine. The user implements risk countermeasures and provides feedback.

[0325] 1. Data Collection and Analysis

[0326] Server Action:

[0327] 1. Data Collection: The server periodically collects risk-related information from internal systems and external data sources, and stores and updates the collected data in a database.

[0328] 2. Data Preprocessing

[0329] Server Action:

[0330] 1. Data preprocessing: The server cleans the collected data, fills in missing data, removes outliers, and then standardizes the data and converts it into a format suitable for training AI models.

[0331] 3. Training the AI ​​model

[0332] Server Action:

[0333] 1. Model training: The server uses the preprocessed data to train an AI model for risk detection, applying machine learning algorithms to learn patterns that identify risks.

[0334] 2. Model evaluation: After training, the server evaluates the accuracy of the model and retrains it if necessary.

[0335] 4. Risk detection and alert generation

[0336] Server Action:

[0337] 1. Real-time data analysis: The server uses AI models to detect risks as business data flows in in real time.

[0338] 2. Risk Scoring: The server calculates a risk score for the detected risks and evaluates their severity.

[0339] 3. Alert Generation: If the risk score exceeds a pre-defined threshold, the server generates a risk alert and notifies the appropriate personnel.

[0340] 5. Alert Notification and Display

[0341] Terminal handling:

[0342] 1. Receiving alert: The terminal receives the risk alert sent from the server.

[0343] 2. Alert display: The device displays detailed information about the risk alert (risk details, scope of impact, recommended countermeasures) to the user.

[0344] 6. Emotion recognition

[0345] Terminal handling:

[0346] 1. Acquiring emotion data: Using the camera and microphone installed on the device, emotions are detected from the user's facial expressions and voice.

[0347] 2. Emotion Analysis: The emotion engine analyzes the emotion data to identify the user's emotional state, for example, to identify the level of stress or anxiety.

[0348] 7. Alert Tuning

[0349] Terminal handling:

[0350] 1. Emotion-based adjustment: The emotion engine adjusts the way risk alerts and recommended actions are displayed based on the user's emotions. For example, if the user is under a lot of stress, more concise and quick actions will be displayed first.

[0351] 8. Implementing risk response measures

[0352] User Action:

[0353] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[0354] 2. Implementing countermeasures: The user implements the necessary risk countermeasures based on the recommended countermeasures displayed, such as suspending transactions, reviewing processes, or launching an internal investigation.

[0355] 9. Providing Feedback

[0356] User Action:

[0357] 1. Providing feedback: Users provide feedback on the results of their risk response measures to the server. The feedback information is used to retrain the AI ​​model, contributing to improving its accuracy.

[0358] 2. Emotional feedback: Providing feedback, including emotional data, on how the user reacted to the alert.

[0359] Specific examples

[0360] As an example, consider a scenario where Company B is managing the risks of a new project. Here are some situations where the emotion engine identifies risks during the project and requires action:

[0361] 1. Data collection: The server collects work data and communication data of project participants.

[0362] 2. Training the AI ​​model: Train the risk detection model based on the pre-processed data.

[0363] 3. Risk detection: The server monitors the progress of the project and calculates a risk score when a risk is detected.

[0364] 4. Alert Generation: The server generates alerts for high-scoring risks and notifies relevant parties.

[0365] 5. Emotion recognition and adjustment: The device detects the project manager's emotions and, if stress levels are high, displays a brief alert to prioritize immediate response.

[0366] 6. Implement risk response measures: The user implements the recommended response measures and feeds the results back to the server.

[0367] 7. Emotional feedback: Feedback containing emotional data is sent to the server and used to improve next risk countermeasures.

[0368] In this way, the present invention enables companies to manage the complex risks they face in real time and utilizes sentiment data to enable more flexible and effective responses.

[0369] The processing flow will be explained below.

[0370] Step 1: Data collection and updates

[0371] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external data sources (e.g., legal databases, SNS). The collected data is stored and updated in a database.

[0372] Step 2: Data Preprocessing

[0373] The server cleans the collected data, fills in missing data, removes outliers, and standardizes (normalizes) the data, converting it into a format suitable for learning AI models, thereby improving the accuracy of model training.

[0374] Step 3: Training the AI ​​model

[0375] The server uses the preprocessed data to train the AI ​​model. Specifically, it uses machine learning algorithms (e.g., random forests, deep learning) to learn patterns to identify risks. After training, the server evaluates the accuracy of the model and retrains it if necessary.

[0376] Step 4: Real-time data analysis

[0377] The server uses AI models to detect risks as business data flows in in real time, using data stream analysis techniques to identify outliers and unexpected patterns.

[0378] Step 5: Risk Scoring

[0379] The server calculates a risk score for each detected risk, based on the scope of impact and probability of occurrence, and serves as a basis for assessing the severity of the risk.

[0380] Step 6: Alert Generation

[0381] The server generates risk alerts when the risk score exceeds a pre-defined threshold, with detailed information about the risk, its scope, and recommended actions to take.

[0382] Step 7: Alert Notifications

[0383] The terminal receives risk alerts sent from the server, and when an alert arrives, it notifies the user so that they can check it immediately.

[0384] Step 8: Displaying alerts

[0385] The device displays detailed information about the risk alert to the user, including the type of risk, the scope of impact, and specific recommended actions to take.

[0386] Step 9: Obtaining Emotion Data

[0387] The device is equipped with a camera and microphone, which are used to capture emotion data from the user's facial expressions and voice. The emotion engine analyzes this data and identifies the user's emotional state.

[0388] Step 10: Adjust alerts based on emotion

[0389] The device adjusts the way risk alerts and recommended actions are displayed based on the user's emotions as recognized by the emotion engine. For example, if the user is under a high level of stress, the device will prioritize displaying concise and specific actions.

[0390] Step 11: Implement risk responses

[0391] Users can check the alerts displayed on their devices and take necessary risk management measures based on the recommended actions presented, such as suspending trading, reviewing processes, or launching an internal investigation.

[0392] Step 12: Provide feedback

[0393] Users then provide feedback on the results of their risk response measures to the server. The feedback information is used to retrain the AI ​​model, helping to improve its accuracy. The feedback also includes user emotional data, which is used to improve future risk detection and response measures.

[0394] Example 2

[0395] 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."

[0396] Conventional risk management systems can detect risks and generate alerts, but they are unable to consider the user's emotional state, making it difficult to present accurate and prompt countermeasures. Furthermore, retraining including user emotional data to improve the accuracy of feedback has not been performed, limiting the improvement of the system's accuracy.

[0397] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0398] In this invention, the server includes means for collecting risk-related information from a data source, means for learning and building a risk detection model based on the collected information, means for detecting risks in real time using the learned risk detection model, means for generating a risk alert based on the detected risk and notifying a responsible person, means for displaying the generated risk alert and recommended countermeasures on a user terminal, means for acquiring and analyzing user emotion data, means for adjusting the display method of the risk alert and recommended countermeasures based on the emotion analysis results, and means for providing feedback on the results of the risk countermeasures implemented based on the risk alert. This enables accurate and prompt presentation of risk countermeasures that take the user's emotional state into consideration, and enables continuous improvement of the accuracy of the system using feedback.

[0399] "Data Sources" are internal and external information sources used to provide risk-related information.

[0400] A "risk detection model" is an AI model that is trained and constructed to identify and evaluate risks based on collected information.

[0401] "Means for detecting risks in real time" refers to a function that uses AI models to instantly identify risks when business data flows in.

[0402] A "Risk Alert" is a notification of a detected risk, including detailed information about the risk and recommended actions to take.

[0403] "Means for notifying the person in charge" refers to a communication means for notifying the person in charge of detected risk information, and includes, for example, email and push notification.

[0404] "User device" means a device used to display risk alerts and recommended countermeasures to a user, including a personal computer or smartphone.

[0405] "Emotion data" is data that indicates the emotional state of the user, obtained from facial expressions, tone of voice, and the like.

[0406] "Emotion analysis" is the process of identifying a user's emotional state based on acquired emotion data.

[0407] The "means for adjusting the display method" is a function for changing the display format of risk alerts and recommended countermeasures based on the results of sentiment analysis.

[0408] A "risk response" is a specific action or procedure taken based on a risk alert to reduce or avoid a risk.

[0409] "Feedback" is the process by which the user sends back to the server the results of the risk response actions they have taken and their emotional state.

[0410] This invention relates to a system for companies to perform efficient and sophisticated risk management. In particular, by combining an emotion engine that recognizes user emotions, it provides a more accurate risk detection and alert system. This system operates between a server, terminals, and users.

[0411] Server Processing

[0412] Data collection and analysis

[0413] The server periodically collects risk-related information from internal systems and external data sources (e.g., the company's ERP system, CRM database, or news feeds and social media). This information is stored in an integrated database and updated on demand for analysis. For example, the server retrieves economic data using APIs every night and imports it into the database.

[0414] Data Preprocessing

[0415] The server cleans the collected data, complements incomplete data, and removes outliers. Next, it standardizes the data and converts it into a format suitable for AI model training. For example, it automatically removes outliers such as "0" and "NULL" from the collected data and inserts average values ​​into missing values.

[0416] Training and evaluating AI models

[0417] The server uses the preprocessed data to train an AI model for risk detection. Specifically, it applies a machine learning algorithm (e.g., random forest) to learn patterns for identifying risks. For example, the server uses the training data to build a random forest model and extracts features of risk factors. After training, the server evaluates the accuracy of the model using test data and retrains it if the accuracy is insufficient. For example, the server evaluates the model's accuracy using cross-validation and retrains the model if the F1 score falls below 80%.

[0418] Risk detection and alert generation

[0419] The server analyzes business data in real time and detects risks using AI models. For example, the server receives new trading data every minute and immediately analyzes it using AI models. It calculates a risk score for the detected risks and evaluates their severity. For example, if the server detects abnormal trading activity, it assigns a risk score to that activity and determines the severity of the risk based on the score. If the risk score exceeds a set threshold, it generates a risk alert and sends a notification to the appropriate person in charge. For example, if the risk score exceeds 80, the server generates a risk alert and sends a notification to the email address of the designated person in charge.

[0420] Terminal handling

[0421] Alert Notification and Display

[0422] The device receives the risk alert sent from the server. For example, the device receives a push notification from the server and the alert is ready to be displayed. Then, the device displays detailed information about the risk alert (risk details, impact scope, and recommended countermeasures) to the user. For example, the device opens a pop-up window and notifies the user, "A new risk has been detected. Please check the impact scope and recommended countermeasures."

[0423] emotion recognition

[0424] The device uses a camera and microphone to acquire emotions from the user's facial expressions and tone of voice. For example, the device uses a camera to capture the user's facial expressions while checking an alert, and records the user's tone of voice with a microphone. The emotion engine analyzes the acquired emotion data and identifies the user's emotional state. For example, the emotion engine uses facial recognition technology to analyze the user's facial expressions and determine that the user is in a "high stress state."

[0425] Alert Tuning

[0426] The emotion engine recognizes the user's emotions and adjusts the way risk alerts and recommended solutions are displayed. For example, if the user is under a lot of stress, the device will prioritize displaying "simple and quick solutions" and postpone detailed explanations.

[0427] User Action

[0428] Implementing risk response measures

[0429] The user checks the details of the risk alert displayed on the terminal and takes necessary risk countermeasures based on the recommended countermeasures displayed. For example, the user reads the details of the alert displayed on the terminal screen, clicks the "pause trading" button, and sends instructions to the system.

[0430] Providing Feedback

[0431] The user feeds back the results of the risk countermeasures they have implemented to the server. The feedback information is used to retrain the AI ​​model. For example, the user inputs whether the implemented countermeasures were successful or not and sends that data to the server. Feedback including the user's emotional data is also provided and used for the next risk countermeasure. For example, the user inputs their emotional state after implementing a countermeasure into their device, and that information is sent to the server and used to improve the AI ​​model.

[0432] Examples and prompts

[0433] For example, consider a scenario where a company is starting a new project and managing the risks associated with it. Here's how the emotion engine might identify risks during the project and prompt a rapid response:

[0434] 1. Data Collection:

[0435] The server collects work and communication data from the project team, for example, it uses APIs to extract data from a project management tool.

[0436] 2. Training the AI ​​model:

[0437] Based on the preprocessed data, a risk detection model is trained using a framework such as TensorFlow. For example, the server uses project data to learn patterns of risk occurrence.

[0438] 3. Risk detection:

[0439] The server monitors the project management system (PMS) in real time, identifying and scoring risks. For example, the server periodically updates data and calculates risk scores in real time.

[0440] 4. Alert Generation:

[0441] If the risk score is high, the server sends a risk alert to the project team's chat tool (e.g., a team communication tool). For example, if a serious risk is detected, the server notifies the team that an "urgent risk has occurred."

[0442] 5. Emotion recognition and regulation:

[0443] The device detects the project manager's emotions and, if stress levels are high, the device will simplify the alert display and prioritize immediate action. For example, the device will display a "summary of countermeasures" and encourage the manager to check the details later.

[0444] 6. Implementing risk response measures:

[0445] The project manager implements the recommended action and sends the results to the server, for example, the project manager clicks the "Stop Transaction" button and records the results.

[0446] 7. Emotional Feedback:

[0447] The project manager reports their emotional state after implementing the countermeasures, and this information is used to plan the next risk countermeasure. For example, the project manager inputs their impression of "stress reduction," and the data is sent to the server.

[0448] Examples of prompt statements

[0449] Below are some example prompts to input to a generative AI model:

[0450] 1. Example prompt 1:

[0451] "Identify risk factors in new projects. Generate a risk score by predicting the likelihood of risk occurrence based on project team work data and historical risk data."

[0452] 2. Example prompt 2:

[0453] "Analyze the emotional state of the project leader and, if the stress level is high, adjust the alert display and suggest quick countermeasures. Capture emotional data using the camera and microphone and analyze it with the emotion engine."

[0454] In this way, the present invention manages the complex risks faced by companies in real time and enables more flexible and effective responses by utilizing user emotional data.

[0455] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0456] System program processing flow and explanation

[0457] Step 1: Data collection

[0458] Server Processing

[0459] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM databases) and external data sources (e.g., news feeds, economic data, social media) using APIs and scraping techniques.

[0460] Input: API call, scraping script

[0461] Output: Risk-related information dataset

[0462] Specific operation: For example, every night at night, the server retrieves economic data using an API and imports it into the database.

[0463] Step 2: Data Preprocessing

[0464] Server Processing

[0465] The server cleans the collected data, fills in incomplete data, removes outliers, and then standardizes the data and converts it into a format suitable for AI model training.

[0466] Input: Risk-related information dataset

[0467] Output: Preprocessed dataset

[0468] Specific operation: For example, automatically remove outliers such as "0" and "NULL" from collected data and insert the average value into missing values.

[0469] Step 3: Training the AI ​​model

[0470] Server Processing

[0471] The server uses the preprocessed data to train an AI model for risk detection, specifically by applying machine learning algorithms (e.g., random forests) to learn patterns to identify risks.

[0472] Input: Preprocessed dataset

[0473] Output: A trained AI model

[0474] Specific operations: For example, the server uses training data to build a random forest model and extract features of risk factors.

[0475] Step 4: Model evaluation

[0476] Server Processing

[0477] After training, the server evaluates the accuracy of the model using test data and retrains it if the accuracy is insufficient.

[0478] Input: trained AI model, test dataset

[0479] Output: Evaluation results, improved AI model if necessary

[0480] Specific behavior: For example, the server evaluates the accuracy of the model using cross-validation and retrains the model if the F1 score falls below 80%.

[0481] Step 5: Real-time data analysis

[0482] Server Processing

[0483] The server analyzes business data in real time and uses AI models to detect risks.

[0484] Input: Real-time business data

[0485] Output: Detected risk information

[0486] Specific operation: For example, the server receives new transaction data every minute and immediately analyzes it using an AI model.

[0487] Step 6: Risk Scoring

[0488] Server Processing

[0489] The server calculates a risk score for the detected risk and evaluates its severity.

[0490] Input: Detected risk information

[0491] Output: Risk score, risk severity rating

[0492] Specific operation: For example, if the server detects abnormal trading activity, it assigns a risk score to the activity and determines the severity of the risk based on the score.

[0493] Step 7: Alert Generation

[0494] Server Processing

[0495] If the risk score exceeds a set threshold, the server generates a risk alert and sends a notification to the appropriate personnel.

[0496] Inputs: Risk score, risk severity rating

[0497] Output: Risk alert notification

[0498] Specific behavior: For example, if the risk score exceeds 80, the server generates a risk alert and sends a notification to the email address of the designated person in charge.

[0499] Step 8: Receive and view alerts

[0500] Terminal handling

[0501] The terminal receives risk alerts sent from the server and displays detailed information about the risk alert (risk details, scope of impact, recommended countermeasures) to the user.

[0502] Input: Risk Alert Notification

[0503] Output: Alert details display

[0504] Specific behavior: For example, the device will open a pop-up window and notify you that "A new risk has been detected. Please check the scope of impact and recommended countermeasures."

[0505] Step 9: Acquire and analyze emotion data

[0506] Terminal handling

[0507] Using the device's built-in camera and microphone, emotions are acquired from the user's facial expressions and tone of voice. The emotion engine analyzes the acquired emotional data and identifies the user's emotional state.

[0508] Input: Camera video, microphone audio

[0509] Output: Emotion analysis results

[0510] Specific operation: For example, the device captures the user's facial expression while checking the alert with a camera and records the tone of voice with a microphone. The emotion engine uses facial recognition technology to analyze the user's facial expression and determine whether they are in a "high stress state."

[0511] Step 10: Adjust alerts based on emotion

[0512] Terminal handling

[0513] Tailor how risk alerts and recommended actions are presented based on the user's emotions as recognized by the emotion engine.

[0514] Input: Sentiment analysis results, risk alerts

[0515] Output: Adjusted alert display

[0516] Specific operation: For example, if the user is under a lot of stress, the device will prioritize displaying "simple and quick solutions" and postpone detailed explanations.

[0517] Step 11: Implement risk responses

[0518] User Action

[0519] The user checks the content of the risk alert displayed on the device and takes the necessary risk countermeasures based on the recommended countermeasures displayed.

[0520] Input: Risk alert, recommended action

[0521] Output: Risk response measures taken, response results

[0522] Specific actions: For example, the user reads the details of the alert displayed on the terminal screen, clicks the "Pause trading" button, and sends instructions to the system.

[0523] Step 12: Provide feedback and retrain the model

[0524] User and Server Processing

[0525] Users provide feedback on the results of their risk countermeasures to the server, which uses the feedback information to retrain the AI ​​model.

[0526] Input: Countermeasure execution results, emotional feedback

[0527] Output: Improved AI model

[0528] Specific operation: For example, the user inputs whether the implemented countermeasure was successful and sends the data to the server. The user also reports their emotional state after the countermeasure was implemented, and this information is sent to the server and used to improve the AI ​​model.

[0529] (Application example 2)

[0530] 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."

[0531] While conventional risk management systems are capable of detecting risks based on data collection and analysis, they lack the flexibility to respond flexibly, taking into account the user's emotional state. This can result in inappropriate adjustment of risk alerts and limited effectiveness in implementing countermeasures. Furthermore, the lack of user emotional feedback leads to insufficient improvement in the accuracy of risk management.

[0532] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0533] In this invention, the server includes: means for collecting risk-related information from a data source; means for learning and building a risk detection model based on the collected information; means for detecting risks in real time using the learned risk detection model; means for generating a risk alert based on the detected risk and notifying a responsible party; means for displaying the generated risk alert and recommended countermeasures on a user terminal; means for recognizing a user's emotions in real time using a camera and microphone mounted on the terminal; means for adjusting the display method of the risk alert based on the emotions recognized by the emotion engine; and means for feeding back the results of the risk countermeasures implemented based on the risk alert. This allows user emotion data to be incorporated into the risk management system, enabling more accurate risk detection and flexible responses.

[0534] "Data Source" means a source of information for collecting risk-related information, including internal systems and external databases.

[0535] A "risk detection model" is an AI model that is trained using machine learning algorithms to identify and assess risks based on collected information.

[0536] "Real-time" refers to the fact that information is processed and analyzed instantly, without any time lag.

[0537] A "risk alert" is a warning notification generated based on a detected risk, which includes details of the risk and recommended countermeasures.

[0538] "User terminal" refers to a device for displaying risk alerts and recommended countermeasures, including smartphones and smart glasses.

[0539] The "emotion engine" is software that uses the device's built-in camera and microphone to recognize and analyze the user's emotional state in real time.

[0540] "Risk score" is a numerical representation of the severity of the risk assessed by the AI ​​model.

[0541] "Feedback" refers to the results of risk response measures taken by the user and emotional data being sent back to the server, and is information used to retrain the AI ​​model.

[0542] "Preprocessing" is the process of cleaning collected data, filling in missing data, removing outliers, standardizing it, and converting it into a format suitable for learning AI models.

[0543] This invention relates to a smart logistics risk management system that improves work efficiency and safety within logistics centers. Specifically, it has the function of recognizing the emotional state of workers in real time through terminals equipped with an emotion engine and adjusting risk alerts. This system operates as follows between the server, terminals, and users.

[0544] Server Processing

[0545] The server handles data collection, data preprocessing, AI model training and retraining, risk detection, risk scoring, and alert generation. Specific software used includes machine learning algorithms and sentiment engines. Data is standardized, and risk detection models are trained using algorithms such as RandomForestClassifier.

[0546] Data collection and preprocessing

[0547] The server periodically collects risk-related information from internal systems and external databases, stores it in a database, and then pre-processes the data by cleaning it, removing outliers, and filling in missing data, converting it into a format suitable for training AI models.

[0548] Training and evaluating AI models

[0549] The risk detection model is trained based on the pre-processed data. After training, the model undergoes accuracy evaluation and is retrained as necessary. Emotion recognition data from the emotion engine is also integrated into the risk detection model, contributing to improving the model's accuracy.

[0550] Terminal handling

[0551] The device will recognize emotions in real time, display and adjust risk alerts, and collect user feedback. The hardware used includes smart glasses and smartphones with an integrated emotion engine.

[0552] Emotion Recognition and Alert Adjustment

[0553] Using cameras and microphones mounted on smart glasses or smartphones, the system detects emotions from the worker's facial expressions and voice. An emotion engine analyzes the emotional data and identifies the user's emotional state. Based on this information, the system adjusts the way risk alerts are displayed. For example, if stress levels are high, quick and concise alerts will be prioritized.

[0554] User Action

[0555] The user checks the alerts displayed on their device, takes recommended risk countermeasures, and then provides the results and emotional feedback to the server, which then retrains the AI ​​model and improves its accuracy.

[0556] Specific examples

[0557] Consider the following scenario as an example for Logistics Center A. The server collects work data and environmental data from cameras and sensors within the center, as well as facial expression and voice data from workers' smart glasses. Based on the pre-processed data, a risk detection model is trained and performs real-time data analysis to detect risks. A risk score indicates the severity, and an appropriate risk alert is generated and notified to the terminal. The terminal recognizes the user's emotional state in real time and adjusts the alert display method. Workers implement the displayed risk countermeasures and provide the results and emotional feedback to the server.

[0558] For example, the following prompt sentences are applied as example input to a generative AI model:

[0559] "A logistics center worker shows signs of high stress while performing tasks. Their facial expression indicates stress with a confidence level of 0.7. The current risk score for the ongoing task is 0.8. Provide a concise alert message."

[0560] This approach streamlines risk management within the distribution center and improves worker safety.

[0561] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0562] Step 1: Data collection

[0563] The server periodically collects risk-related information from internal systems and external databases and stores it in a database. Specific inputs include work data, environmental data, and communication data. This ensures the data necessary for training risk detection models and real-time analysis. The output is organized raw data stored in the database.

[0564] Step 2: Data Preprocessing

[0565] The server cleans the collected data, fills in missing data, and removes outliers. It then standardizes the data and converts it into a format suitable for AI model training. Specific operations include data standardization, outlier elimination, and missing value imputation. The input is raw data, and the output is preprocessed data.

[0566] Step 3: Training the AI ​​model

[0567] The server uses the preprocessed data to train an AI model for risk detection. It applies a machine learning algorithm (e.g., RandomForestClassifier) ​​to learn patterns to identify risks. After training, it evaluates the accuracy of the model and retrains it if necessary. The input is the preprocessed data, and the output is a trained risk detection model.

[0568] Step 4: Obtaining emotion recognition data

[0569] The device uses the on-board camera and microphone to recognize emotions from the user's facial expressions and voice in real time. The emotion engine analyzes the emotion data and identifies the user's emotional state. Specific operations include face detection, voice analysis, and emotion classification. The input is raw data acquired by the camera and microphone, and the output is analyzed emotion data.

[0570] Step 5: Risk detection and alert generation

[0571] The server analyzes incoming data in real time and detects risks using a trained risk detection model. It then calculates a risk score and evaluates the severity. If the risk score exceeds a set threshold, it generates a risk alert and notifies the appropriate personnel. The input is real-time data, and the output is the generated risk alert.

[0572] Step 6: Adjust alerts based on emotion

[0573] The device adjusts the display of risk alerts and recommended countermeasures based on the user's emotions identified by the emotion engine. For example, if the user is in a state of high stress, it prioritizes displaying simpler and quicker countermeasures. The input is the analyzed emotion data and the generated risk alert, and the output is the adjusted risk alert display.

[0574] Step 7: Implement risk responses

[0575] The user checks the risk alert displayed on the terminal and implements the recommended countermeasures. Specific actions include changing work methods, pausing processes, contacting other personnel, etc. The input is the adjusted risk alert, and the output is feedback on the implementation of the countermeasures.

[0576] Step 8: Provide feedback

[0577] The user provides the server with the results of the risk countermeasures they have implemented and their emotional feedback. The server then retrains the AI ​​model based on this feedback data to improve the model's accuracy. The input is the user's feedback data, and the output is the retrained AI model.

[0578] For example, the following prompt sentences are applied as example input to a generative AI model:

[0579] "A logistics center worker shows signs of high stress while performing tasks. Their facial expression indicates stress with a confidence level of 0.7. The current risk score for the ongoing task is 0.8. Provide a concise alert message."

[0580] 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.

[0581] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (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.

[0582] 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.

[0583] [Second embodiment]

[0584] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0585] 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.

[0586] 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).

[0587] 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.

[0588] 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.

[0589] 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).

[0590] 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.

[0591] 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.

[0592] 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.

[0593] 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.

[0594] 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.

[0595] 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."

[0596] The present invention relates to a system for enabling a company to efficiently manage risk. The specific processing content of the program of this system and an embodiment thereof will be described below with specific examples.

[0597] Overall flow

[0598] This system operates between the server, terminals, and users as follows: The server collects and analyzes data, trains AI models, and detects risks and generates alerts. The terminals are responsible for displaying alert notifications and risk information, while users implement risk countermeasures and provide feedback.

[0599] 1. Data Collection and Analysis

[0600] Server Action:

[0601] 1. Data collection: The server collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external legal databases and social networking sites.

[0602] 2. Data Update: The server saves and updates the collected data to the database used by date and time.

[0603] 2. Learning and training the AI ​​model

[0604] Server Action:

[0605] 1. Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the AI ​​model.

[0606] 2. Model training: The server uses the pre-processed data to train an AI model for risk detection, leveraging machine learning algorithms (e.g., random forest, deep learning).

[0607] 3. Model evaluation: The server evaluates the accuracy of the model and retrains it if necessary.

[0608] 3. Risk detection and alert generation

[0609] Server Action:

[0610] 1. Real-time data analysis: The server analyzes incoming business data in real time and detects risks using AI models.

[0611] 2. Risk Scoring: The server calculates a risk score for each detected risk and evaluates its severity.

[0612] 3. Alert generation: If the risk score exceeds a certain threshold, the server generates an alert and notifies the appropriate personnel.

[0613] 4. Alert notification and display

[0614] Terminal handling:

[0615] 1. Receiving an alert: The terminal receives an alert sent from the server.

[0616] 2. Alert display: The terminal displays detailed information about the alert (risk details, scope of impact, recommended countermeasures) to the user.

[0617] 5. Risk Management

[0618] User Action:

[0619] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[0620] 2. Implementing countermeasures: The user implements the necessary risk countermeasures based on the recommended countermeasures presented, such as suspending transactions, reviewing processes, or initiating an internal investigation.

[0621] 3. Providing feedback: The user provides feedback on the results of the implementation to the server, and this information is used to retrain the model.

[0622] Specific examples

[0623] As an example, consider a scenario in which Company A detects reputation risks, including negative reactions on social media, when releasing a new product.

[0624] 1. Data Collection: The server collects social media data and internal marketing data.

[0625] 2. Training the AI ​​model: Train the risk detection model based on the pre-processed data.

[0626] 3. Risk detection: When a new product is released, the server detects negative feedback in real time and calculates a high risk score.

[0627] 4. Alert Generation: The server generates and notifies the marketing department personnel of a risk alert.

[0628] 5. Notification and response: The device receives an alert, the user checks the relevant information, and modifies the new marketing strategy based on the recommendations.

[0629] In this way, the present invention enables efficient management of the complex risks faced by companies and realizes prompt and appropriate responses.

[0630] The processing flow will be explained below.

[0631] Step 1: Data collection and updates

[0632] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external data sources (e.g., legal databases, SNS). The collected data is stored and updated in a database.

[0633] Step 2: Data Preprocessing

[0634] The server cleans the collected data, fills in missing data, removes outliers, and then standardizes (normalizes) the data and converts it into a format suitable for training AI models.

[0635] Step 3: Training the AI ​​model

[0636] The server uses the preprocessed data to train the AI ​​model, applying machine learning algorithms to learn patterns to identify risks. After training, the model's accuracy is evaluated and retrained if necessary.

[0637] Step 4: Real-time data analysis

[0638] The server uses AI models to detect risks as business data flows in in real time, using data stream analysis techniques to identify outliers and unexpected patterns.

[0639] Step 5: Risk Scoring

[0640] The server calculates a risk score for each detected risk, based on the scope of impact and probability of occurrence, and serves as a basis for assessing the severity of the risk.

[0641] Step 6: Alert Generation

[0642] The server generates risk alerts when the risk score exceeds a pre-defined threshold, with detailed information about the risk, its scope, and recommended actions to take.

[0643] Step 7: Alert Notifications

[0644] The terminal receives risk alerts sent from the server, and when an alert arrives, it notifies the user so that they can check it immediately.

[0645] Step 8: Displaying alerts

[0646] The device displays detailed information about the risk alert to the user, including the type of risk, the scope of impact, and specific recommended actions to take.

[0647] Step 9: Implement risk responses

[0648] Based on the displayed risk alerts and recommended actions, users can take necessary risk response measures, such as suspending trading, reviewing processes, or launching an internal investigation.

[0649] Step 10: Provide feedback

[0650] Users then provide feedback on the results of their risk response measures to the server, which uses the feedback information to subsequently retrain the AI ​​model, contributing to improving the model's accuracy.

[0651] Example 1

[0652] 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."

[0653] Companies are being asked to improve the efficiency of risk management and respond quickly, but many challenges exist in collecting information from diverse data sources, training AI models, detecting and notifying risks in real time, and incorporating feedback. In particular, the processes of data preprocessing and model retraining are complex, and integrating the entire system takes time and effort. Furthermore, real-time risk detection and accurate alert generation are difficult, which can lead to a lack of speed and accuracy in risk responses. It is desirable to resolve these issues.

[0654] 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.

[0655] In this invention, the server includes a means for collecting risk-related information from data sources, a means for storing and updating the collected information in a database by date and time, and a means for preprocessing the collected data and converting it into a format suitable for training a generative AI model, thereby enabling companies to efficiently manage risk using an integrated system.

[0656] "Data Source" is a collective term for internal systems and external sources used to collect risk-related information.

[0657] A "risk detection model" is a general term for algorithms and machine learning models that are learned and built based on collected data to detect risks.

[0658] "Real-time" refers to processing and analysis being carried out immediately, close to the moment data is generated.

[0659] "Risk alert" refers to a warning or caution message that is generated based on a detected risk and notified to a responsible person.

[0660] "User Terminal" refers collectively to any electronic device used to display risk alerts and recommended responses.

[0661] "Feedback" refers to providing information about the results of risk response measures to the server.

[0662] A "generative AI model" refers to a model built using machine learning algorithms based on collected data.

[0663] "Preprocessing" is the process of organizing and transforming collected data to convert it into a form suitable for training an AI model.

[0664] An "algorithm" refers to a combination of steps or computational methods for solving a specific problem.

[0665] A "hyperparameter" is a setting value during the training process of a machine learning model, and refers to an element that affects the performance of the model.

[0666] A "stream processing framework" refers to a software platform for processing and analyzing data in real time.

[0667] The present invention relates to a system for enabling a company to efficiently manage risk. The specific processing contents of the program of this system and an embodiment thereof will be described below.

[0668] Overall flow

[0669] This system operates between the server, terminals, and users as follows: The server collects and analyzes data, trains generative AI models, and detects risks and generates alerts. The terminals are responsible for notifying alerts and displaying risk information, while users implement risk countermeasures and provide feedback.

[0670] Data collection and analysis

[0671] Server Processing

[0672] 1. Data collection: The server obtains business data from internal systems (e.g., ERP systems, CRM systems) and also collects risk-related information from external legal databases and social media APIs.

[0673] The server obtains financial and supply chain data by calling the ERP system's API.

[0674] Use a social media API (e.g., Twitter API) to collect tweets related to a specific hashtag.

[0675] Use a web scraping tool (e.g., BeautifulSoup) to scrape new legal information from the legal database.

[0676] 2. Data update: The server saves and updates the collected data in a database (e.g., MySQL, PostgreSQL) according to the date and time.

[0677] The new data is stored in a temporary table and then reconciled with the existing database using a journal file to ensure data consistency.

[0678] If the data is updated, update the timestamp and insert the new data.

[0679] Learning and training AI models

[0680] Server Processing

[0681] 1. Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the generative AI model, specifically by imputing missing data, converting character data to numeric data, and normalizing the data.

[0682] Missing data will be imputed using the median.

[0683] Character data is converted to numeric data using one-hot encoding.

[0684] Normalization is performed to align the range of the data.

[0685] 2. Model training: The preprocessed data is used to train a generative AI model. Model training uses machine learning frameworks such as TensorFlow or scikit-learn, and employs algorithms such as random forests and deep learning.

[0686] The dataset is divided into batches for training.

[0687] Tune the hyperparameters of your model to get optimal performance.

[0688] 3. Model evaluation: Evaluate the accuracy of the trained model and retrain it if necessary.

[0689] For evaluation, we use a test dataset and calculate precision, recall, and F1 score.

[0690] If the accuracy is low, reset the hyperparameters and retrain.

[0691] Risk detection and alert generation

[0692] Server Processing

[0693] 1. Real-time data analysis: The server analyzes incoming business data in real time and uses generative AI models to detect risks, using stream processing frameworks such as Apache Kafka and Apache Flink.

[0694] Monitor real-time data streams and detect specific patterns.

[0695] Input data into the AI ​​model to obtain a predicted value.

[0696] 2. Risk Scoring: The server calculates a risk score for each detected risk and evaluates its severity.

[0697] Set scoring criteria according to the type of risk.

[0698] A score is calculated and evaluated based on the output value of the AI ​​model.

[0699] 3. Alert generation: If the risk score exceeds the set threshold, the server generates an alert and notifies the responsible person.

[0700] An email containing the alert content is created and sent using the SAP communication protocol.

[0701] Use a push notification service to send alerts to the mobile devices of personnel.

[0702] Alert Notification and Display

[0703] Terminal handling

[0704] 1. Receiving an alert: The device receives an alert sent from the server.

[0705] Receives push notifications from the notification server and displays them on the device.

[0706] 2. Alert display: Displays detailed information about the received alert (risk content, scope of impact, recommended countermeasures) to the user.

[0707] The mobile or desktop app will pop up an alert message.

[0708] Risk Management

[0709] User Action

[0710] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[0711] Check the alert displayed on the device screen in detail.

[0712] 2. Implementing Actions: The user implements the necessary risk response measures based on the recommended actions, which may include suspending trading, reviewing processes, or initiating an internal investigation.

[0713] Implement countermeasures based on company policy.

[0714] 3. Providing feedback: The results of the implemented countermeasures are provided to the server as feedback.

[0715] Feedback data is sent to the server using a dedicated input form.

[0716] Specific examples

[0717] Consider a scenario in which Company A detects reputation risks, including negative reactions on social media, when releasing a new product.

[0718] 1. Data Collection: The server collects social media data and internal marketing data.

[0719] 2. Training a generative AI model: Train a risk detection model based on the preprocessed data.

[0720] 3. Risk detection: When a new product is released, the server detects negative feedback in real time and calculates a high risk score.

[0721] 4. Alert generation: The server generates and notifies the marketing department personnel of a risk alert.

[0722] 5. Notification and response: The device receives an alert, the user checks the relevant information, and then a new marketing strategy is modified based on the recommendations.

[0723] Example prompt: "Describe a system that detects negative reactions on social media in real time when a company releases a new product and notifies the marketing department."

[0724] In this way, the present invention enables efficient management of the complex risks faced by companies and realizes prompt and appropriate responses.

[0725] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0726] Step 1: Data collection

[0727] The server retrieves business data from internal systems (e.g., ERP systems, CRM systems) and also collects risk-related information using external legal databases and SNS APIs.

[0728] Input: Business data, legal data, and social media data from data sources.

[0729] Specific behavior:

[0730] The server obtains financial and supply chain data by calling the ERP system's API.

[0731] Use a social media API (e.g., Twitter API) to collect tweets related to a specific hashtag.

[0732] Use a web scraping tool (e.g., BeautifulSoup) to scrape new legal information from the legal database.

[0733] Output: The retrieved data.

[0734] Step 2: Data Update

[0735] The server stores and updates the collected data in a database (e.g., MySQL, PostgreSQL) according to the date and time.

[0736] Input: The retrieved data.

[0737] Specific behavior:

[0738] Save the new data in a temporary table and compare it with the existing database to ensure data consistency during the update.

[0739] When data is updated, the timestamp is also updated and the latest data is saved.

[0740] Output: The updated database.

[0741] Step 3: Data Preprocessing

[0742] The server preprocesses the collected data and converts it into a format suitable for training the generative AI model, specifically by imputing missing data, converting character data to numeric data, and normalizing the data.

[0743] Input: Data from the updated database.

[0744] Specific behavior:

[0745] Missing data will be imputed using the median.

[0746] Character data is converted to numeric data using one-hot encoding.

[0747] Normalize the data to the range 0 to 1.

[0748] Output: Preprocessed data.

[0749] Step 4: Model training

[0750] The preprocessed data is used to train generative AI models, using machine learning frameworks including TensorFlow and scikit-learn.

[0751] Input: Preprocessed data.

[0752] Specific behavior:

[0753] Define a dataset and train batch by batch.

[0754] Set and tune the hyperparameters of your model for optimal performance.

[0755] Deep learning algorithms are used to train the models.

[0756] Output: A trained AI model.

[0757] Step 5: Model evaluation

[0758] Evaluate the accuracy of trained AI models and retrain them if necessary.

[0759] Input: A trained AI model, a test dataset.

[0760] Specific behavior:

[0761] The model is evaluated using a test dataset.

[0762] Calculate precision, recall, and F1 score to measure the performance of the model.

[0763] Adjust hyperparameters and retrain as needed.

[0764] Output: The evaluated AI model.

[0765] Step 6: Real-time data analysis

[0766] The server uses stream processing frameworks such as Apache Kafka and Apache Flink to analyze incoming business data in real time and detect risks using generative AI models.

[0767] Input: Business data flowing in in real time.

[0768] Specific behavior:

[0769] Monitor real-time data streams and detect important patterns.

[0770] The data is analyzed using an AI model to obtain predicted values.

[0771] Output: Risk detection results.

[0772] Step 7: Risk Scoring

[0773] The server calculates a risk score for each detected risk and evaluates its severity.

[0774] Input: Risk findings.

[0775] Specific behavior:

[0776] Use pre-defined scoring criteria for each type of risk.

[0777] A risk score is calculated based on the output value of the AI ​​model.

[0778] Output: Risk scoring results.

[0779] Step 8: Alert Generation

[0780] If the risk score exceeds a set threshold, the server generates an alert and notifies the responsible person.

[0781] Input: Risk scoring results.

[0782] Specific behavior:

[0783] Create and send an email containing the alert content.

[0784] Send alerts to personnel's mobile devices via push notification services.

[0785] Output: The generated alert.

[0786] Step 9: Receiving and viewing alerts

[0787] The terminal receives the alert sent from the server and displays the detailed information to the user.

[0788] Input: The generated alert.

[0789] Specific behavior:

[0790] Receives push notifications from the notification server and displays them on the device.

[0791] Pop up an alert message in the mobile or desktop app.

[0792] Output: The displayed alert.

[0793] Step 10: Risk response

[0794] The user checks the contents of the alert displayed on the terminal and implements the necessary risk countermeasures based on the recommended countermeasures.

[0795] Input: The displayed alert.

[0796] Specific behavior:

[0797] Check the alert displayed on the device screen.

[0798] In accordance with our internal policies, we will take appropriate action, which may include suspending business, reviewing processes, or initiating an internal investigation.

[0799] Output: The results of the implemented measures.

[0800] Step 11: Provide feedback

[0801] The user provides the results of the risk countermeasures implemented to the server as feedback.

[0802] Input: The result of the action taken.

[0803] Specific behavior:

[0804] Feedback data is sent to the server using a dedicated input form.

[0805] The server stores the feedback data and uses it to improve the model in the future.

[0806] Output: The newly collected feedback data.

[0807] (Application example 1)

[0808] 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."

[0809] With the evolution of autonomous vehicles, it has become increasingly important to adapt to real-time changes in traffic conditions and the surrounding environment and quickly and appropriately detect and manage risks. However, with conventional technology, it has been difficult to efficiently analyze the large amounts of data obtained from sensors and cameras and appropriately evaluate and notify risks. This has resulted in delayed responses to risks such as traffic accidents and breakdowns, which could result in serious damage. Therefore, improvements to real-time risk management systems are required.

[0810] 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.

[0811] In this invention, the server includes means for collecting risk-related information from a data source, means for learning and building a risk detection model based on the collected information, means for detecting risks in real time using the learned risk detection model, means for generating a risk alert based on the detected risk and notifying a responsible person, means for displaying the generated risk alert and recommended countermeasures on a user terminal, means for feeding back the results of executing the risk countermeasures based on the risk alert, means for collecting data in real time from sensors and cameras of the autonomous vehicle and analyzing traffic conditions and the surrounding environment, and means for displaying and executing countermeasures in the autonomous vehicle based on the risk alert, thereby enabling real-time risk management of the autonomous vehicle.

[0812] "Data sources" are sources of information used to collect risk-related information, such as internal systems, external legal databases, and social media.

[0813] A "risk detection model" is a trained AI model for detecting risks based on collected data, using machine learning algorithms such as random forest and deep learning.

[0814] A "risk alert" is a warning message that is generated based on a detected risk and is sent to a responsible person.

[0815] "User terminal" refers to a device that displays alert notifications and risk information. Specifically, this applies to smartphones and head-mounted displays.

[0816] A "sensor" is a device installed in an autonomous vehicle that measures speed, distance, acceleration, etc.

[0817] The "camera" is an image capture device that visualizes the surrounding environment of an autonomous vehicle and is used to grasp the situation in real time.

[0818] "Traffic conditions" refers to information that comprehensively indicates the position, speed, direction, etc. of vehicles and pedestrians on the road.

[0819] "Surrounding environment" refers to the physical and dynamic elements surrounding an autonomous vehicle, including buildings, other vehicles, and pedestrians.

[0820] "Feedback" is information used to evaluate the results of risk response measures taken and to help with the next model training.

[0821] "Real-time" refers to the near-instantaneous collection, analysis, notification, and response of information.

[0822] "Recommended countermeasures" refers to specific countermeasures or action plans suggested to users in response to detected risks.

[0823] The present invention relates to a system that collects and analyzes data in real time from sensors and cameras in autonomous vehicles, and quickly and appropriately detects and manages risks according to traffic conditions and the surrounding environment. Specific embodiments for carrying out the present invention are described below.

[0824] This system mainly consists of three components: a server, a terminal, and a user.

[0825] server

[0826] The server is responsible for data collection, analysis, training of AI models, risk detection, and alert generation. Specifically, it has the following functions:

[0827] Data collection: The server collects data in real time from the autonomous vehicle's sensors and cameras. The hardware used here includes the autonomous vehicle's speed sensor, distance sensor, acceleration sensor, camera, etc.

[0828] Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the AI ​​model. Preprocessing includes data normalization and feature engineering. Software used includes Python, Pandas, and Numpy.

[0829] Training the AI ​​model: The server uses the preprocessed data to train a risk detection model. Machine learning algorithms include random forests and deep learning. Software used here includes scikit-learn.

[0830] Risk detection: Analyzes and detects risks in real time using trained AI models, and if a risk is detected, calculates a risk score and assesses its severity.

[0831] Alert generation: When a risk exceeds a certain threshold in importance, a risk alert is generated and the responsible person is notified.

[0832] Terminal

[0833] The device receives risk alerts from the server and notifies the user. Specifically, it has the following functions:

[0834] Receive alerts: Receive risk alerts sent from the server in real time.

[0835] Alert display: Received alerts are displayed visually and intuitively to the user on devices such as smartphones and head-mounted displays. These devices display detailed information about the alert (risk details, scope of impact, recommended countermeasures).

[0836] User

[0837] Users receive risk alerts via their devices and take appropriate countermeasures, which are detailed below:

[0838] Alert confirmation: The user confirms the content of the risk alert displayed on the device.

[0839] Countermeasure Execution: Based on the proposed countermeasures, implement the necessary risk countermeasures, such as applying emergency braking or proposing a route change.

[0840] Providing feedback: The results of the implemented countermeasures are fed back to the server. This information is used for the next retraining of the AI ​​model.

[0841] Examples and prompts

[0842] Examples:

[0843] When an autonomous vehicle is traveling on a highway, sensors detect a sudden decrease in the distance to the vehicle ahead. This data is pre-processed, and an AI model evaluates the risk in real time. If the risk is deemed high, an alert is generated. The alert is immediately displayed on a smartphone or head-mounted display, prompting the driver to apply emergency braking.

[0844] Prompt statement:

[0845] Write code to assess risk and generate an alert if the risk is high, based on the following autonomous vehicle sensor data:

[0846] Speed: float

[0847] Proximity: float

[0848] Acceleration: float

[0849] Include data pre-processing, training advanced AI models, data analysis, and alert generation as needed.

[0850] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0851] Step 1:

[0852] Data collection

[0853] The server collects real-time data from the autonomous vehicle's sensors and cameras, including speed, proximity, acceleration, etc. The collected data serves as input for subsequent processing steps.

[0854] Step 2:

[0855] Data Preprocessing

[0856] The server preprocesses the data collected in step 1. This preprocessing involves normalizing the data and performing feature engineering. For example, it normalizes the speed data to fit within a specified range and removes outliers. This preprocessed data serves as input for training and evaluating the AI ​​model.

[0857] Step 3:

[0858] Training an AI model

[0859] The server trains an AI model based on the preprocessed data from step 2. It uses algorithms such as random forest and deep learning to build a risk detection model. This trained model generates output results for detecting risks in real time.

[0860] Step 4:

[0861] Real-time risk detection

[0862] The server uses a trained AI model to analyze the data collected in real time and detect risks. For example, a sudden decrease in the distance ahead of the vehicle is considered a high risk. The results of this analysis serve as input data for subsequent alert generation.

[0863] Step 5:

[0864] Risk Scoring and Alerting

[0865] The server calculates a risk score and evaluates the severity based on the risks detected in step 4. If the risk score exceeds a certain threshold, it generates a risk alert and notifies the responsible party. The alert includes details of the risk, the scope of impact, and recommended countermeasures.

[0866] Step 6:

[0867] Alert Notification and Display

[0868] The device receives risk alerts sent from the server and notifies the user. The alerts are displayed via a user interface such as a smartphone or head-mounted display. Recommendations for risk countermeasures are also displayed on the device in real time.

[0869] Step 7:

[0870] Implementing risk response measures

[0871] The user checks the risk alerts and recommended countermeasures displayed on the device and takes specific countermeasures against the risk, such as applying emergency braking or proposing a route change. The results of this action become input data for the next feedback step.

[0872] Step 8:

[0873] Providing Feedback

[0874] Users can provide feedback on the results of their risk countermeasures to the server, which will be used to retrain the AI ​​model and improve the accuracy of the risk detection model.

[0875] 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.

[0876] The present invention relates to a system for companies to perform efficient and sophisticated risk management. In particular, by combining it with an emotion engine that recognizes user emotions, it provides a more accurate risk detection and alert system.

[0877] Overall flow

[0878] This system operates as follows between the server, terminal, and user: The server collects and analyzes data, trains AI models, detects risks, and generates alerts. The terminal is responsible for notifying alerts and displaying risk information, and recognizes user emotions using an emotion engine. The user implements risk countermeasures and provides feedback.

[0879] 1. Data Collection and Analysis

[0880] Server Action:

[0881] 1. Data Collection: The server periodically collects risk-related information from internal systems and external data sources, and stores and updates the collected data in a database.

[0882] 2. Data Preprocessing

[0883] Server Action:

[0884] 1. Data preprocessing: The server cleans the collected data, fills in missing data, removes outliers, and then standardizes the data and converts it into a format suitable for training AI models.

[0885] 3. Training the AI ​​model

[0886] Server Action:

[0887] 1. Model training: The server uses the preprocessed data to train an AI model for risk detection, applying machine learning algorithms to learn patterns that identify risks.

[0888] 2. Model evaluation: After training, the server evaluates the accuracy of the model and retrains it if necessary.

[0889] 4. Risk detection and alert generation

[0890] Server Action:

[0891] 1. Real-time data analysis: The server uses AI models to detect risks as business data flows in in real time.

[0892] 2. Risk Scoring: The server calculates a risk score for the detected risks and evaluates their severity.

[0893] 3. Alert Generation: If the risk score exceeds a pre-defined threshold, the server generates a risk alert and notifies the appropriate personnel.

[0894] 5. Alert Notification and Display

[0895] Terminal handling:

[0896] 1. Receiving alert: The terminal receives the risk alert sent from the server.

[0897] 2. Alert display: The device displays detailed information about the risk alert (risk details, scope of impact, recommended countermeasures) to the user.

[0898] 6. Emotion recognition

[0899] Terminal handling:

[0900] 1. Acquiring emotion data: Using the camera and microphone installed on the device, emotions are detected from the user's facial expressions and voice.

[0901] 2. Emotion Analysis: The emotion engine analyzes the emotion data to identify the user's emotional state, for example, to identify the level of stress or anxiety.

[0902] 7. Alert Tuning

[0903] Terminal handling:

[0904] 1. Emotion-based adjustment: The emotion engine adjusts the way risk alerts and recommended actions are displayed based on the user's emotions. For example, if the user is under a lot of stress, more concise and quick actions will be displayed first.

[0905] 8. Implementing risk response measures

[0906] User Action:

[0907] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[0908] 2. Implementing countermeasures: The user implements the necessary risk countermeasures based on the recommended countermeasures displayed, such as suspending transactions, reviewing processes, or launching an internal investigation.

[0909] 9. Providing Feedback

[0910] User Action:

[0911] 1. Providing feedback: Users provide feedback on the results of their risk response measures to the server. The feedback information is used to retrain the AI ​​model, contributing to improving its accuracy.

[0912] 2. Emotional feedback: Providing feedback, including emotional data, on how the user reacted to the alert.

[0913] Specific examples

[0914] As an example, consider a scenario where Company B is managing the risks of a new project. Here are some situations where the emotion engine identifies risks during the project and requires action:

[0915] 1. Data collection: The server collects work data and communication data of project participants.

[0916] 2. Training the AI ​​model: Train the risk detection model based on the pre-processed data.

[0917] 3. Risk detection: The server monitors the progress of the project and calculates a risk score when a risk is detected.

[0918] 4. Alert Generation: The server generates alerts for high-scoring risks and notifies relevant parties.

[0919] 5. Emotion recognition and adjustment: The device detects the project manager's emotions and, if stress levels are high, displays a brief alert to prioritize immediate response.

[0920] 6. Implement risk response measures: The user implements the recommended response measures and feeds the results back to the server.

[0921] 7. Emotional feedback: Feedback containing emotional data is sent to the server and used to improve next risk countermeasures.

[0922] In this way, the present invention enables companies to manage the complex risks they face in real time and utilizes sentiment data to enable more flexible and effective responses.

[0923] The processing flow will be explained below.

[0924] Step 1: Data collection and updates

[0925] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external data sources (e.g., legal databases, SNS). The collected data is stored and updated in a database.

[0926] Step 2: Data Preprocessing

[0927] The server cleans the collected data, fills in missing data, removes outliers, and standardizes (normalizes) the data, converting it into a format suitable for learning AI models, thereby improving the accuracy of model training.

[0928] Step 3: Training the AI ​​model

[0929] The server uses the preprocessed data to train the AI ​​model. Specifically, it uses machine learning algorithms (e.g., random forests, deep learning) to learn patterns to identify risks. After training, the server evaluates the accuracy of the model and retrains it if necessary.

[0930] Step 4: Real-time data analysis

[0931] The server uses AI models to detect risks as business data flows in in real time, using data stream analysis techniques to identify outliers and unexpected patterns.

[0932] Step 5: Risk Scoring

[0933] The server calculates a risk score for each detected risk, based on the scope of impact and probability of occurrence, and serves as a basis for assessing the severity of the risk.

[0934] Step 6: Alert Generation

[0935] The server generates risk alerts when the risk score exceeds a pre-defined threshold, with detailed information about the risk, its scope, and recommended actions to take.

[0936] Step 7: Alert Notifications

[0937] The terminal receives risk alerts sent from the server, and when an alert arrives, it notifies the user so that they can check it immediately.

[0938] Step 8: Displaying alerts

[0939] The device displays detailed information about the risk alert to the user, including the type of risk, the scope of impact, and specific recommended actions to take.

[0940] Step 9: Obtaining Emotion Data

[0941] The device is equipped with a camera and microphone, which are used to capture emotion data from the user's facial expressions and voice. The emotion engine analyzes this data and identifies the user's emotional state.

[0942] Step 10: Adjust alerts based on emotion

[0943] The device adjusts the way risk alerts and recommended actions are displayed based on the user's emotions as recognized by the emotion engine. For example, if the user is under a high level of stress, the device will prioritize displaying concise and specific actions.

[0944] Step 11: Implement risk responses

[0945] Users can check the alerts displayed on their devices and take necessary risk management measures based on the recommended actions presented, such as suspending trading, reviewing processes, or launching an internal investigation.

[0946] Step 12: Provide feedback

[0947] Users then provide feedback on the results of their risk response measures to the server. The feedback information is used to retrain the AI ​​model, helping to improve its accuracy. The feedback also includes user emotional data, which is used to improve future risk detection and response measures.

[0948] Example 2

[0949] 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."

[0950] Conventional risk management systems can detect risks and generate alerts, but they are unable to consider the user's emotional state, making it difficult to present accurate and prompt countermeasures. Furthermore, retraining including user emotional data to improve the accuracy of feedback has not been performed, limiting the improvement of the system's accuracy.

[0951] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0952] In this invention, the server includes means for collecting risk-related information from a data source, means for learning and building a risk detection model based on the collected information, means for detecting risks in real time using the learned risk detection model, means for generating a risk alert based on the detected risk and notifying a responsible person, means for displaying the generated risk alert and recommended countermeasures on a user terminal, means for acquiring and analyzing user emotion data, means for adjusting the display method of the risk alert and recommended countermeasures based on the emotion analysis results, and means for providing feedback on the results of the risk countermeasures implemented based on the risk alert. This enables accurate and prompt presentation of risk countermeasures that take the user's emotional state into consideration, and enables continuous improvement of the accuracy of the system using feedback.

[0953] "Data Sources" are internal and external information sources used to provide risk-related information.

[0954] A "risk detection model" is an AI model that is trained and constructed to identify and evaluate risks based on collected information.

[0955] "Means for detecting risks in real time" refers to a function that uses AI models to instantly identify risks when business data flows in.

[0956] A "Risk Alert" is a notification of a detected risk, including detailed information about the risk and recommended actions to take.

[0957] "Means for notifying the person in charge" refers to a communication means for notifying the person in charge of detected risk information, and includes, for example, email and push notification.

[0958] "User device" means a device used to display risk alerts and recommended countermeasures to a user, including a personal computer or smartphone.

[0959] "Emotion data" is data that indicates the emotional state of the user, obtained from facial expressions, tone of voice, and the like.

[0960] "Emotion analysis" is the process of identifying a user's emotional state based on acquired emotion data.

[0961] The "means for adjusting the display method" is a function for changing the display format of risk alerts and recommended countermeasures based on the results of sentiment analysis.

[0962] A "risk response" is a specific action or procedure taken based on a risk alert to reduce or avoid a risk.

[0963] "Feedback" is the process by which the user sends back to the server the results of the risk response actions they have taken and their emotional state.

[0964] This invention relates to a system for companies to perform efficient and sophisticated risk management. In particular, by combining an emotion engine that recognizes user emotions, it provides a more accurate risk detection and alert system. This system operates between a server, terminals, and users.

[0965] Server Processing

[0966] Data collection and analysis

[0967] The server periodically collects risk-related information from internal systems and external data sources (e.g., the company's ERP system, CRM database, or news feeds and social media). This information is stored in an integrated database and updated on demand for analysis. For example, the server retrieves economic data using APIs every night and imports it into the database.

[0968] Data Preprocessing

[0969] The server cleans the collected data, complements incomplete data, and removes outliers. Next, it standardizes the data and converts it into a format suitable for AI model training. For example, it automatically removes outliers such as "0" and "NULL" from the collected data and inserts average values ​​into missing values.

[0970] Training and evaluating AI models

[0971] The server uses the preprocessed data to train an AI model for risk detection. Specifically, it applies a machine learning algorithm (e.g., random forest) to learn patterns for identifying risks. For example, the server uses the training data to build a random forest model and extracts features of risk factors. After training, the server evaluates the accuracy of the model using test data and retrains it if the accuracy is insufficient. For example, the server evaluates the model's accuracy using cross-validation and retrains the model if the F1 score falls below 80%.

[0972] Risk detection and alert generation

[0973] The server analyzes business data in real time and detects risks using AI models. For example, the server receives new trading data every minute and immediately analyzes it using AI models. It calculates a risk score for the detected risks and evaluates their severity. For example, if the server detects abnormal trading activity, it assigns a risk score to that activity and determines the severity of the risk based on the score. If the risk score exceeds a set threshold, it generates a risk alert and sends a notification to the appropriate person in charge. For example, if the risk score exceeds 80, the server generates a risk alert and sends a notification to the email address of the designated person in charge.

[0974] Terminal handling

[0975] Alert Notification and Display

[0976] The device receives the risk alert sent from the server. For example, the device receives a push notification from the server and the alert is ready to be displayed. Then, the device displays detailed information about the risk alert (risk details, impact scope, and recommended countermeasures) to the user. For example, the device opens a pop-up window and notifies the user, "A new risk has been detected. Please check the impact scope and recommended countermeasures."

[0977] emotion recognition

[0978] The device uses a camera and microphone to acquire emotions from the user's facial expressions and tone of voice. For example, the device uses a camera to capture the user's facial expressions while checking an alert, and records the user's tone of voice with a microphone. The emotion engine analyzes the acquired emotion data and identifies the user's emotional state. For example, the emotion engine uses facial recognition technology to analyze the user's facial expressions and determine that the user is in a "high stress state."

[0979] Alert Tuning

[0980] The emotion engine recognizes the user's emotions and adjusts the way risk alerts and recommended solutions are displayed. For example, if the user is under a lot of stress, the device will prioritize displaying "simple and quick solutions" and postpone detailed explanations.

[0981] User Action

[0982] Implementing risk response measures

[0983] The user checks the details of the risk alert displayed on the terminal and takes necessary risk countermeasures based on the recommended countermeasures displayed. For example, the user reads the details of the alert displayed on the terminal screen, clicks the "pause trading" button, and sends instructions to the system.

[0984] Providing Feedback

[0985] The user feeds back the results of the risk countermeasures they have implemented to the server. The feedback information is used to retrain the AI ​​model. For example, the user inputs whether the implemented countermeasures were successful or not and sends that data to the server. Feedback including the user's emotional data is also provided and used for the next risk countermeasure. For example, the user inputs their emotional state after implementing a countermeasure into their device, and that information is sent to the server and used to improve the AI ​​model.

[0986] Examples and prompts

[0987] For example, consider a scenario where a company is starting a new project and managing the risks associated with it. Here's how the emotion engine might identify risks during the project and prompt a rapid response:

[0988] 1. Data Collection:

[0989] The server collects work and communication data from the project team, for example, it uses APIs to extract data from a project management tool.

[0990] 2. Training the AI ​​model:

[0991] Based on the preprocessed data, a risk detection model is trained using a framework such as TensorFlow. For example, the server uses project data to learn patterns of risk occurrence.

[0992] 3. Risk detection:

[0993] The server monitors the project management system (PMS) in real time, identifying and scoring risks. For example, the server periodically updates data and calculates risk scores in real time.

[0994] 4. Alert Generation:

[0995] If the risk score is high, the server sends a risk alert to the project team's chat tool (e.g., a team communication tool). For example, if a serious risk is detected, the server notifies the team that an "urgent risk has occurred."

[0996] 5. Emotion recognition and regulation:

[0997] The device detects the project manager's emotions and, if stress levels are high, the device will simplify the alert display and prioritize immediate action. For example, the device will display a "summary of countermeasures" and encourage the manager to check the details later.

[0998] 6. Implementing risk response measures:

[0999] The project manager implements the recommended action and sends the results to the server, for example, the project manager clicks the "Stop Transaction" button and records the results.

[1000] 7. Emotional Feedback:

[1001] The project manager reports their emotional state after implementing the countermeasures, and this information is used to plan the next risk countermeasure. For example, the project manager inputs their impression of "stress reduction," and the data is sent to the server.

[1002] Examples of prompt statements

[1003] Below are some example prompts to input to a generative AI model:

[1004] 1. Example prompt 1:

[1005] "Identify risk factors in new projects. Generate a risk score by predicting the likelihood of risk occurrence based on project team work data and historical risk data."

[1006] 2. Example prompt 2:

[1007] "Analyze the emotional state of the project leader and, if the stress level is high, adjust the alert display and suggest quick countermeasures. Capture emotional data using the camera and microphone and analyze it with the emotion engine."

[1008] In this way, the present invention manages the complex risks faced by companies in real time and enables more flexible and effective responses by utilizing user emotional data.

[1009] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1010] System program processing flow and explanation

[1011] Step 1: Data collection

[1012] Server Processing

[1013] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM databases) and external data sources (e.g., news feeds, economic data, social media) using APIs and scraping techniques.

[1014] Input: API call, scraping script

[1015] Output: Risk-related information dataset

[1016] Specific operation: For example, every night at night, the server retrieves economic data using an API and imports it into the database.

[1017] Step 2: Data Preprocessing

[1018] Server Processing

[1019] The server cleans the collected data, fills in incomplete data, removes outliers, and then standardizes the data and converts it into a format suitable for AI model training.

[1020] Input: Risk-related information dataset

[1021] Output: Preprocessed dataset

[1022] Specific operation: For example, automatically remove outliers such as "0" and "NULL" from collected data and insert the average value into missing values.

[1023] Step 3: Training the AI ​​model

[1024] Server Processing

[1025] The server uses the preprocessed data to train an AI model for risk detection, specifically by applying machine learning algorithms (e.g., random forests) to learn patterns to identify risks.

[1026] Input: Preprocessed dataset

[1027] Output: A trained AI model

[1028] Specific operations: For example, the server uses training data to build a random forest model and extract features of risk factors.

[1029] Step 4: Model evaluation

[1030] Server Processing

[1031] After training, the server evaluates the accuracy of the model using test data and retrains it if the accuracy is insufficient.

[1032] Input: trained AI model, test dataset

[1033] Output: Evaluation results, improved AI model if necessary

[1034] Specific behavior: For example, the server evaluates the accuracy of the model using cross-validation and retrains the model if the F1 score falls below 80%.

[1035] Step 5: Real-time data analysis

[1036] Server Processing

[1037] The server analyzes business data in real time and uses AI models to detect risks.

[1038] Input: Real-time business data

[1039] Output: Detected risk information

[1040] Specific operation: For example, the server receives new transaction data every minute and immediately analyzes it using an AI model.

[1041] Step 6: Risk Scoring

[1042] Server Processing

[1043] The server calculates a risk score for the detected risk and evaluates its severity.

[1044] Input: Detected risk information

[1045] Output: Risk score, risk severity rating

[1046] Specific operation: For example, if the server detects abnormal trading activity, it assigns a risk score to the activity and determines the severity of the risk based on the score.

[1047] Step 7: Alert Generation

[1048] Server Processing

[1049] If the risk score exceeds a set threshold, the server generates a risk alert and sends a notification to the appropriate personnel.

[1050] Inputs: Risk score, risk severity rating

[1051] Output: Risk alert notification

[1052] Specific behavior: For example, if the risk score exceeds 80, the server generates a risk alert and sends a notification to the email address of the designated person in charge.

[1053] Step 8: Receive and view alerts

[1054] Terminal handling

[1055] The terminal receives risk alerts sent from the server and displays detailed information about the risk alert (risk details, scope of impact, recommended countermeasures) to the user.

[1056] Input: Risk Alert Notification

[1057] Output: Alert details display

[1058] Specific behavior: For example, the device will open a pop-up window and notify you that "A new risk has been detected. Please check the scope of impact and recommended countermeasures."

[1059] Step 9: Acquire and analyze emotion data

[1060] Terminal handling

[1061] Using the device's built-in camera and microphone, emotions are acquired from the user's facial expressions and tone of voice. The emotion engine analyzes the acquired emotional data and identifies the user's emotional state.

[1062] Input: Camera video, microphone audio

[1063] Output: Emotion analysis results

[1064] Specific operation: For example, the device captures the user's facial expression while checking the alert with a camera and records the tone of voice with a microphone. The emotion engine uses facial recognition technology to analyze the user's facial expression and determine whether they are in a "high stress state."

[1065] Step 10: Adjust alerts based on emotion

[1066] Terminal handling

[1067] Tailor how risk alerts and recommended actions are presented based on the user's emotions as recognized by the emotion engine.

[1068] Input: Sentiment analysis results, risk alerts

[1069] Output: Adjusted alert display

[1070] Specific operation: For example, if the user is under a lot of stress, the device will prioritize displaying "simple and quick solutions" and postpone detailed explanations.

[1071] Step 11: Implement risk responses

[1072] User Action

[1073] The user checks the content of the risk alert displayed on the device and takes the necessary risk countermeasures based on the recommended countermeasures displayed.

[1074] Input: Risk alert, recommended action

[1075] Output: Risk response measures taken, response results

[1076] Specific actions: For example, the user reads the details of the alert displayed on the terminal screen, clicks the "Pause trading" button, and sends instructions to the system.

[1077] Step 12: Provide feedback and retrain the model

[1078] User and Server Processing

[1079] Users provide feedback on the results of their risk countermeasures to the server, which uses the feedback information to retrain the AI ​​model.

[1080] Input: Countermeasure execution results, emotional feedback

[1081] Output: Improved AI model

[1082] Specific operation: For example, the user inputs whether the implemented countermeasure was successful and sends the data to the server. The user also reports their emotional state after the countermeasure was implemented, and this information is sent to the server and used to improve the AI ​​model.

[1083] (Application example 2)

[1084] 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."

[1085] While conventional risk management systems are capable of detecting risks based on data collection and analysis, they lack the flexibility to respond flexibly, taking into account the user's emotional state. This can result in inappropriate adjustment of risk alerts and limited effectiveness in implementing countermeasures. Furthermore, the lack of user emotional feedback leads to insufficient improvement in the accuracy of risk management.

[1086] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1087] In this invention, the server includes: means for collecting risk-related information from a data source; means for learning and building a risk detection model based on the collected information; means for detecting risks in real time using the learned risk detection model; means for generating a risk alert based on the detected risk and notifying a responsible party; means for displaying the generated risk alert and recommended countermeasures on a user terminal; means for recognizing a user's emotions in real time using a camera and microphone mounted on the terminal; means for adjusting the display method of the risk alert based on the emotions recognized by the emotion engine; and means for feeding back the results of the risk countermeasures implemented based on the risk alert. This allows user emotion data to be incorporated into the risk management system, enabling more accurate risk detection and flexible responses.

[1088] "Data Source" means a source of information for collecting risk-related information, including internal systems and external databases.

[1089] A "risk detection model" is an AI model that is trained using machine learning algorithms to identify and assess risks based on collected information.

[1090] "Real-time" refers to the fact that information is processed and analyzed instantly, without any time lag.

[1091] A "risk alert" is a warning notification generated based on a detected risk, which includes details of the risk and recommended countermeasures.

[1092] "User terminal" refers to a device for displaying risk alerts and recommended countermeasures, including smartphones and smart glasses.

[1093] The "emotion engine" is software that uses the device's built-in camera and microphone to recognize and analyze the user's emotional state in real time.

[1094] "Risk score" is a numerical representation of the severity of the risk assessed by the AI ​​model.

[1095] "Feedback" refers to the results of risk response measures taken by the user and emotional data being sent back to the server, and is information used to retrain the AI ​​model.

[1096] "Preprocessing" is the process of cleaning collected data, filling in missing data, removing outliers, standardizing it, and converting it into a format suitable for learning AI models.

[1097] This invention relates to a smart logistics risk management system that improves work efficiency and safety within logistics centers. Specifically, it has the function of recognizing the emotional state of workers in real time through terminals equipped with an emotion engine and adjusting risk alerts. This system operates as follows between the server, terminals, and users.

[1098] Server Processing

[1099] The server handles data collection, data preprocessing, AI model training and retraining, risk detection, risk scoring, and alert generation. Specific software used includes machine learning algorithms and sentiment engines. Data is standardized, and risk detection models are trained using algorithms such as RandomForestClassifier.

[1100] Data collection and preprocessing

[1101] The server periodically collects risk-related information from internal systems and external databases, stores it in a database, and then pre-processes the data by cleaning it, removing outliers, and filling in missing data, converting it into a format suitable for training AI models.

[1102] Training and evaluating AI models

[1103] The risk detection model is trained based on the pre-processed data. After training, the model undergoes accuracy evaluation and is retrained as necessary. Emotion recognition data from the emotion engine is also integrated into the risk detection model, contributing to improving the model's accuracy.

[1104] Terminal handling

[1105] The device will recognize emotions in real time, display and adjust risk alerts, and collect user feedback. The hardware used includes smart glasses and smartphones with an integrated emotion engine.

[1106] Emotion Recognition and Alert Adjustment

[1107] Using cameras and microphones mounted on smart glasses or smartphones, the system detects emotions from the worker's facial expressions and voice. An emotion engine analyzes the emotional data and identifies the user's emotional state. Based on this information, the system adjusts the way risk alerts are displayed. For example, if stress levels are high, quick and concise alerts will be prioritized.

[1108] User Action

[1109] The user checks the alerts displayed on their device, takes recommended risk countermeasures, and then provides the results and emotional feedback to the server, which then retrains the AI ​​model and improves its accuracy.

[1110] Specific examples

[1111] Consider the following scenario as an example for Logistics Center A. The server collects work data and environmental data from cameras and sensors within the center, as well as facial expression and voice data from workers' smart glasses. Based on the pre-processed data, a risk detection model is trained and performs real-time data analysis to detect risks. A risk score indicates the severity, and an appropriate risk alert is generated and notified to the terminal. The terminal recognizes the user's emotional state in real time and adjusts the alert display method. Workers implement the displayed risk countermeasures and provide the results and emotional feedback to the server.

[1112] For example, the following prompt sentences are applied as example input to a generative AI model:

[1113] "A logistics center worker shows signs of high stress while performing tasks. Their facial expression indicates stress with a confidence level of 0.7. The current risk score for the ongoing task is 0.8. Provide a concise alert message."

[1114] This approach streamlines risk management within the distribution center and improves worker safety.

[1115] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1116] Step 1: Data collection

[1117] The server periodically collects risk-related information from internal systems and external databases and stores it in a database. Specific inputs include work data, environmental data, and communication data. This ensures the data necessary for training risk detection models and real-time analysis. The output is organized raw data stored in the database.

[1118] Step 2: Data Preprocessing

[1119] The server cleans the collected data, fills in missing data, and removes outliers. It then standardizes the data and converts it into a format suitable for AI model training. Specific operations include data standardization, outlier elimination, and missing value imputation. The input is raw data, and the output is preprocessed data.

[1120] Step 3: Training the AI ​​model

[1121] The server uses the preprocessed data to train an AI model for risk detection. It applies a machine learning algorithm (e.g., RandomForestClassifier) ​​to learn patterns to identify risks. After training, it evaluates the accuracy of the model and retrains it if necessary. The input is the preprocessed data, and the output is a trained risk detection model.

[1122] Step 4: Obtaining emotion recognition data

[1123] The device uses the on-board camera and microphone to recognize emotions from the user's facial expressions and voice in real time. The emotion engine analyzes the emotion data and identifies the user's emotional state. Specific operations include face detection, voice analysis, and emotion classification. The input is raw data acquired by the camera and microphone, and the output is analyzed emotion data.

[1124] Step 5: Risk detection and alert generation

[1125] The server analyzes incoming data in real time and detects risks using a trained risk detection model. It then calculates a risk score and evaluates the severity. If the risk score exceeds a set threshold, it generates a risk alert and notifies the appropriate personnel. The input is real-time data, and the output is the generated risk alert.

[1126] Step 6: Adjust alerts based on emotion

[1127] The device adjusts the display of risk alerts and recommended countermeasures based on the user's emotions identified by the emotion engine. For example, if the user is in a state of high stress, it prioritizes displaying simpler and quicker countermeasures. The input is the analyzed emotion data and the generated risk alert, and the output is the adjusted risk alert display.

[1128] Step 7: Implement risk responses

[1129] The user checks the risk alert displayed on the terminal and implements the recommended countermeasures. Specific actions include changing work methods, pausing processes, contacting other personnel, etc. The input is the adjusted risk alert, and the output is feedback on the implementation of the countermeasures.

[1130] Step 8: Provide feedback

[1131] The user provides the server with the results of the risk countermeasures they have implemented and their emotional feedback. The server then retrains the AI ​​model based on this feedback data to improve the model's accuracy. The input is the user's feedback data, and the output is the retrained AI model.

[1132] For example, the following prompt sentences are applied as example input to a generative AI model:

[1133] "A logistics center worker shows signs of high stress while performing tasks. Their facial expression indicates stress with a confidence level of 0.7. The current risk score for the ongoing task is 0.8. Provide a concise alert message."

[1134] 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.

[1135] 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.

[1136] 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.

[1137] [Third embodiment]

[1138] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1139] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[1140] 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).

[1141] 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.

[1142] 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.

[1143] 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).

[1144] 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.

[1145] 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.

[1146] 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.

[1147] 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.

[1148] 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.

[1149] 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."

[1150] The present invention relates to a system for enabling a company to efficiently manage risk. The specific processing content of the program of this system and an embodiment thereof will be described below with specific examples.

[1151] Overall flow

[1152] This system operates between the server, terminals, and users as follows: The server collects and analyzes data, trains AI models, and detects risks and generates alerts. The terminals are responsible for displaying alert notifications and risk information, while users implement risk countermeasures and provide feedback.

[1153] 1. Data Collection and Analysis

[1154] Server Action:

[1155] 1. Data collection: The server collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external legal databases and social networking sites.

[1156] 2. Data Update: The server saves and updates the collected data to the database used by date and time.

[1157] 2. Learning and training the AI ​​model

[1158] Server Action:

[1159] 1. Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the AI ​​model.

[1160] 2. Model training: The server uses the pre-processed data to train an AI model for risk detection, leveraging machine learning algorithms (e.g., random forest, deep learning).

[1161] 3. Model evaluation: The server evaluates the accuracy of the model and retrains it if necessary.

[1162] 3. Risk detection and alert generation

[1163] Server Action:

[1164] 1. Real-time data analysis: The server analyzes incoming business data in real time and detects risks using AI models.

[1165] 2. Risk Scoring: The server calculates a risk score for each detected risk and evaluates its severity.

[1166] 3. Alert generation: If the risk score exceeds a certain threshold, the server generates an alert and notifies the appropriate personnel.

[1167] 4. Alert notification and display

[1168] Terminal handling:

[1169] 1. Receiving an alert: The terminal receives an alert sent from the server.

[1170] 2. Alert display: The terminal displays detailed information about the alert (risk details, scope of impact, recommended countermeasures) to the user.

[1171] 5. Risk Management

[1172] User Action:

[1173] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[1174] 2. Implementing countermeasures: The user implements the necessary risk countermeasures based on the recommended countermeasures presented, such as suspending transactions, reviewing processes, or initiating an internal investigation.

[1175] 3. Providing feedback: The user provides feedback on the results of the implementation to the server, and this information is used to retrain the model.

[1176] Specific examples

[1177] As an example, consider a scenario in which Company A detects reputation risks, including negative reactions on social media, when releasing a new product.

[1178] 1. Data Collection: The server collects social media data and internal marketing data.

[1179] 2. Training the AI ​​model: Train the risk detection model based on the pre-processed data.

[1180] 3. Risk detection: When a new product is released, the server detects negative feedback in real time and calculates a high risk score.

[1181] 4. Alert Generation: The server generates and notifies the marketing department personnel of a risk alert.

[1182] 5. Notification and response: The device receives an alert, the user checks the relevant information, and modifies the new marketing strategy based on the recommendations.

[1183] In this way, the present invention enables efficient management of the complex risks faced by companies and realizes prompt and appropriate responses.

[1184] The processing flow will be explained below.

[1185] Step 1: Data collection and updates

[1186] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external data sources (e.g., legal databases, SNS). The collected data is stored and updated in a database.

[1187] Step 2: Data Preprocessing

[1188] The server cleans the collected data, fills in missing data, removes outliers, and then standardizes (normalizes) the data and converts it into a format suitable for training AI models.

[1189] Step 3: Training the AI ​​model

[1190] The server uses the preprocessed data to train the AI ​​model, applying machine learning algorithms to learn patterns to identify risks. After training, the model's accuracy is evaluated and retrained if necessary.

[1191] Step 4: Real-time data analysis

[1192] The server uses AI models to detect risks as business data flows in in real time, using data stream analysis techniques to identify outliers and unexpected patterns.

[1193] Step 5: Risk Scoring

[1194] The server calculates a risk score for each detected risk, based on the scope of impact and probability of occurrence, and serves as a basis for assessing the severity of the risk.

[1195] Step 6: Alert Generation

[1196] The server generates risk alerts when the risk score exceeds a pre-defined threshold, with detailed information about the risk, its scope, and recommended actions to take.

[1197] Step 7: Alert Notifications

[1198] The terminal receives risk alerts sent from the server, and when an alert arrives, it notifies the user so that they can check it immediately.

[1199] Step 8: Displaying alerts

[1200] The device displays detailed information about the risk alert to the user, including the type of risk, the scope of impact, and specific recommended actions to take.

[1201] Step 9: Implement risk responses

[1202] Based on the displayed risk alerts and recommended actions, users can take necessary risk response measures, such as suspending trading, reviewing processes, or launching an internal investigation.

[1203] Step 10: Provide feedback

[1204] Users then provide feedback on the results of their risk response measures to the server, which uses the feedback information to subsequently retrain the AI ​​model, contributing to improving the model's accuracy.

[1205] Example 1

[1206] 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."

[1207] Companies are being asked to improve the efficiency of risk management and respond quickly, but many challenges exist in collecting information from diverse data sources, training AI models, detecting and notifying risks in real time, and incorporating feedback. In particular, the processes of data preprocessing and model retraining are complex, and integrating the entire system takes time and effort. Furthermore, real-time risk detection and accurate alert generation are difficult, which can lead to a lack of speed and accuracy in risk responses. It is desirable to resolve these issues.

[1208] 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.

[1209] In this invention, the server includes a means for collecting risk-related information from data sources, a means for storing and updating the collected information in a database by date and time, and a means for preprocessing the collected data and converting it into a format suitable for training a generative AI model, thereby enabling companies to efficiently manage risk using an integrated system.

[1210] "Data Source" is a collective term for internal systems and external sources used to collect risk-related information.

[1211] A "risk detection model" is a general term for algorithms and machine learning models that are learned and built based on collected data to detect risks.

[1212] "Real-time" refers to processing and analysis being carried out immediately, close to the moment data is generated.

[1213] "Risk alert" refers to a warning or caution message that is generated based on a detected risk and notified to a responsible person.

[1214] "User Terminal" refers collectively to any electronic device used to display risk alerts and recommended responses.

[1215] "Feedback" refers to providing information about the results of risk response measures to the server.

[1216] A "generative AI model" refers to a model built using machine learning algorithms based on collected data.

[1217] "Preprocessing" is the process of organizing and transforming collected data to convert it into a form suitable for training an AI model.

[1218] An "algorithm" refers to a combination of steps or computational methods for solving a specific problem.

[1219] A "hyperparameter" is a setting value during the training process of a machine learning model, and refers to an element that affects the performance of the model.

[1220] A "stream processing framework" refers to a software platform for processing and analyzing data in real time.

[1221] The present invention relates to a system for enabling a company to efficiently manage risk. The specific processing contents of the program of this system and an embodiment thereof will be described below.

[1222] Overall flow

[1223] This system operates between the server, terminals, and users as follows: The server collects and analyzes data, trains generative AI models, and detects risks and generates alerts. The terminals are responsible for notifying alerts and displaying risk information, while users implement risk countermeasures and provide feedback.

[1224] Data collection and analysis

[1225] Server Processing

[1226] 1. Data collection: The server obtains business data from internal systems (e.g., ERP systems, CRM systems) and also collects risk-related information from external legal databases and social media APIs.

[1227] The server obtains financial and supply chain data by calling the ERP system's API.

[1228] Use a social media API (e.g., Twitter API) to collect tweets related to a specific hashtag.

[1229] Use a web scraping tool (e.g., BeautifulSoup) to scrape new legal information from the legal database.

[1230] 2. Data update: The server saves and updates the collected data in a database (e.g., MySQL, PostgreSQL) according to the date and time.

[1231] The new data is stored in a temporary table and then reconciled with the existing database using a journal file to ensure data consistency.

[1232] If the data is updated, update the timestamp and insert the new data.

[1233] Learning and training AI models

[1234] Server Processing

[1235] 1. Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the generative AI model, specifically by imputing missing data, converting character data to numeric data, and normalizing the data.

[1236] Missing data will be imputed using the median.

[1237] Character data is converted to numeric data using one-hot encoding.

[1238] Normalization is performed to align the range of the data.

[1239] 2. Model training: The preprocessed data is used to train a generative AI model. Model training uses machine learning frameworks such as TensorFlow or scikit-learn, and employs algorithms such as random forests and deep learning.

[1240] The dataset is divided into batches for training.

[1241] Tune the hyperparameters of your model to get optimal performance.

[1242] 3. Model evaluation: Evaluate the accuracy of the trained model and retrain it if necessary.

[1243] For evaluation, we use a test dataset and calculate precision, recall, and F1 score.

[1244] If the accuracy is low, reset the hyperparameters and retrain.

[1245] Risk detection and alert generation

[1246] Server Processing

[1247] 1. Real-time data analysis: The server analyzes incoming business data in real time and uses generative AI models to detect risks, using stream processing frameworks such as Apache Kafka and Apache Flink.

[1248] Monitor real-time data streams and detect specific patterns.

[1249] Input data into the AI ​​model to obtain a predicted value.

[1250] 2. Risk Scoring: The server calculates a risk score for each detected risk and evaluates its severity.

[1251] Set scoring criteria according to the type of risk.

[1252] A score is calculated and evaluated based on the output value of the AI ​​model.

[1253] 3. Alert generation: If the risk score exceeds the set threshold, the server generates an alert and notifies the responsible person.

[1254] An email containing the alert content is created and sent using the SAP communication protocol.

[1255] Use a push notification service to send alerts to the mobile devices of personnel.

[1256] Alert Notification and Display

[1257] Terminal handling

[1258] 1. Receiving an alert: The device receives an alert sent from the server.

[1259] Receives push notifications from the notification server and displays them on the device.

[1260] 2. Alert display: Displays detailed information about the received alert (risk content, scope of impact, recommended countermeasures) to the user.

[1261] The mobile or desktop app will pop up an alert message.

[1262] Risk Management

[1263] User Action

[1264] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[1265] Check the alert displayed on the device screen in detail.

[1266] 2. Implementing Actions: The user implements the necessary risk response measures based on the recommended actions, which may include suspending trading, reviewing processes, or initiating an internal investigation.

[1267] Implement countermeasures based on company policy.

[1268] 3. Providing feedback: The results of the implemented countermeasures are provided to the server as feedback.

[1269] Feedback data is sent to the server using a dedicated input form.

[1270] Specific examples

[1271] Consider a scenario in which Company A detects reputation risks, including negative reactions on social media, when releasing a new product.

[1272] 1. Data Collection: The server collects social media data and internal marketing data.

[1273] 2. Training a generative AI model: Train a risk detection model based on the preprocessed data.

[1274] 3. Risk detection: When a new product is released, the server detects negative feedback in real time and calculates a high risk score.

[1275] 4. Alert generation: The server generates and notifies the marketing department personnel of a risk alert.

[1276] 5. Notification and response: The device receives an alert, the user checks the relevant information, and then a new marketing strategy is modified based on the recommendations.

[1277] Example prompt: "Describe a system that detects negative reactions on social media in real time when a company releases a new product and notifies the marketing department."

[1278] In this way, the present invention enables efficient management of the complex risks faced by companies and realizes prompt and appropriate responses.

[1279] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1280] Step 1: Data collection

[1281] The server retrieves business data from internal systems (e.g., ERP systems, CRM systems) and also collects risk-related information using external legal databases and SNS APIs.

[1282] Input: Business data, legal data, and social media data from data sources.

[1283] Specific behavior:

[1284] The server obtains financial and supply chain data by calling the ERP system's API.

[1285] Use a social media API (e.g., Twitter API) to collect tweets related to a specific hashtag.

[1286] Use a web scraping tool (e.g., BeautifulSoup) to scrape new legal information from the legal database.

[1287] Output: The retrieved data.

[1288] Step 2: Data Update

[1289] The server stores and updates the collected data in a database (e.g., MySQL, PostgreSQL) according to the date and time.

[1290] Input: The retrieved data.

[1291] Specific behavior:

[1292] Save the new data in a temporary table and compare it with the existing database to ensure data consistency during the update.

[1293] When data is updated, the timestamp is also updated and the latest data is saved.

[1294] Output: The updated database.

[1295] Step 3: Data Preprocessing

[1296] The server preprocesses the collected data and converts it into a format suitable for training the generative AI model, specifically by imputing missing data, converting character data to numeric data, and normalizing the data.

[1297] Input: Data from the updated database.

[1298] Specific behavior:

[1299] Missing data will be imputed using the median.

[1300] Character data is converted to numeric data using one-hot encoding.

[1301] Normalize the data to the range 0 to 1.

[1302] Output: Preprocessed data.

[1303] Step 4: Model training

[1304] The preprocessed data is used to train generative AI models, using machine learning frameworks including TensorFlow and scikit-learn.

[1305] Input: Preprocessed data.

[1306] Specific behavior:

[1307] Define a dataset and train batch by batch.

[1308] Set and tune the hyperparameters of your model for optimal performance.

[1309] Deep learning algorithms are used to train the models.

[1310] Output: A trained AI model.

[1311] Step 5: Model evaluation

[1312] Evaluate the accuracy of trained AI models and retrain them if necessary.

[1313] Input: A trained AI model, a test dataset.

[1314] Specific behavior:

[1315] The model is evaluated using a test dataset.

[1316] Calculate precision, recall, and F1 score to measure the performance of the model.

[1317] Adjust hyperparameters and retrain as needed.

[1318] Output: The evaluated AI model.

[1319] Step 6: Real-time data analysis

[1320] The server uses stream processing frameworks such as Apache Kafka and Apache Flink to analyze incoming business data in real time and detect risks using generative AI models.

[1321] Input: Business data flowing in in real time.

[1322] Specific behavior:

[1323] Monitor real-time data streams and detect important patterns.

[1324] The data is analyzed using an AI model to obtain predicted values.

[1325] Output: Risk detection results.

[1326] Step 7: Risk Scoring

[1327] The server calculates a risk score for each detected risk and evaluates its severity.

[1328] Input: Risk findings.

[1329] Specific behavior:

[1330] Use pre-defined scoring criteria for each type of risk.

[1331] A risk score is calculated based on the output value of the AI ​​model.

[1332] Output: Risk scoring results.

[1333] Step 8: Alert Generation

[1334] If the risk score exceeds a set threshold, the server generates an alert and notifies the responsible person.

[1335] Input: Risk scoring results.

[1336] Specific behavior:

[1337] Create and send an email containing the alert content.

[1338] Send alerts to personnel's mobile devices via push notification services.

[1339] Output: The generated alert.

[1340] Step 9: Receiving and viewing alerts

[1341] The terminal receives the alert sent from the server and displays the detailed information to the user.

[1342] Input: The generated alert.

[1343] Specific behavior:

[1344] Receives push notifications from the notification server and displays them on the device.

[1345] Pop up an alert message in the mobile or desktop app.

[1346] Output: The displayed alert.

[1347] Step 10: Risk response

[1348] The user checks the contents of the alert displayed on the terminal and implements the necessary risk countermeasures based on the recommended countermeasures.

[1349] Input: The displayed alert.

[1350] Specific behavior:

[1351] Check the alert displayed on the device screen.

[1352] In accordance with our internal policies, we will take appropriate action, which may include suspending business, reviewing processes, or initiating an internal investigation.

[1353] Output: The results of the implemented measures.

[1354] Step 11: Provide feedback

[1355] The user provides the results of the risk countermeasures implemented to the server as feedback.

[1356] Input: The result of the action taken.

[1357] Specific behavior:

[1358] Feedback data is sent to the server using a dedicated input form.

[1359] The server stores the feedback data and uses it to improve the model in the future.

[1360] Output: The newly collected feedback data.

[1361] (Application example 1)

[1362] 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."

[1363] With the evolution of autonomous vehicles, it has become increasingly important to adapt to real-time changes in traffic conditions and the surrounding environment and quickly and appropriately detect and manage risks. However, with conventional technology, it has been difficult to efficiently analyze the large amounts of data obtained from sensors and cameras and appropriately evaluate and notify risks. This has resulted in delayed responses to risks such as traffic accidents and breakdowns, which could result in serious damage. Therefore, improvements to real-time risk management systems are required.

[1364] 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.

[1365] In this invention, the server includes means for collecting risk-related information from a data source, means for learning and building a risk detection model based on the collected information, means for detecting risks in real time using the learned risk detection model, means for generating a risk alert based on the detected risk and notifying a responsible person, means for displaying the generated risk alert and recommended countermeasures on a user terminal, means for feeding back the results of executing the risk countermeasures based on the risk alert, means for collecting data in real time from sensors and cameras of the autonomous vehicle and analyzing traffic conditions and the surrounding environment, and means for displaying and executing countermeasures in the autonomous vehicle based on the risk alert, thereby enabling real-time risk management of the autonomous vehicle.

[1366] "Data sources" are sources of information used to collect risk-related information, such as internal systems, external legal databases, and social media.

[1367] A "risk detection model" is a trained AI model for detecting risks based on collected data, using machine learning algorithms such as random forest and deep learning.

[1368] A "risk alert" is a warning message that is generated based on a detected risk and is sent to a responsible person.

[1369] "User terminal" refers to a device that displays alert notifications and risk information. Specifically, this applies to smartphones and head-mounted displays.

[1370] A "sensor" is a device installed in an autonomous vehicle that measures speed, distance, acceleration, etc.

[1371] The "camera" is an image capture device that visualizes the surrounding environment of an autonomous vehicle and is used to grasp the situation in real time.

[1372] "Traffic conditions" refers to information that comprehensively indicates the position, speed, direction, etc. of vehicles and pedestrians on the road.

[1373] "Surrounding environment" refers to the physical and dynamic elements surrounding an autonomous vehicle, including buildings, other vehicles, and pedestrians.

[1374] "Feedback" is information used to evaluate the results of risk response measures taken and to help with the next model training.

[1375] "Real-time" refers to the near-instantaneous collection, analysis, notification, and response of information.

[1376] "Recommended countermeasures" refers to specific countermeasures or action plans suggested to users in response to detected risks.

[1377] The present invention relates to a system that collects and analyzes data in real time from sensors and cameras in autonomous vehicles, and quickly and appropriately detects and manages risks according to traffic conditions and the surrounding environment. Specific embodiments for carrying out the present invention are described below.

[1378] This system mainly consists of three components: a server, a terminal, and a user.

[1379] server

[1380] The server is responsible for data collection, analysis, training of AI models, risk detection, and alert generation. Specifically, it has the following functions:

[1381] Data collection: The server collects data in real time from the autonomous vehicle's sensors and cameras. The hardware used here includes the autonomous vehicle's speed sensor, distance sensor, acceleration sensor, camera, etc.

[1382] Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the AI ​​model. Preprocessing includes data normalization and feature engineering. Software used includes Python, Pandas, and Numpy.

[1383] Training the AI ​​model: The server uses the preprocessed data to train a risk detection model. Machine learning algorithms include random forests and deep learning. Software used here includes scikit-learn.

[1384] Risk detection: Analyzes and detects risks in real time using trained AI models, and if a risk is detected, calculates a risk score and assesses its severity.

[1385] Alert generation: When a risk exceeds a certain threshold in importance, a risk alert is generated and the responsible person is notified.

[1386] Terminal

[1387] The device receives risk alerts from the server and notifies the user. Specifically, it has the following functions:

[1388] Receive alerts: Receive risk alerts sent from the server in real time.

[1389] Alert display: Received alerts are displayed visually and intuitively to the user on devices such as smartphones and head-mounted displays. These devices display detailed information about the alert (risk details, scope of impact, recommended countermeasures).

[1390] User

[1391] Users receive risk alerts via their devices and take appropriate countermeasures, which are detailed below:

[1392] Alert confirmation: The user confirms the content of the risk alert displayed on the device.

[1393] Countermeasure Execution: Based on the proposed countermeasures, implement the necessary risk countermeasures, such as applying emergency braking or proposing a route change.

[1394] Providing feedback: The results of the implemented countermeasures are fed back to the server. This information is used for the next retraining of the AI ​​model.

[1395] Examples and prompts

[1396] Examples:

[1397] When an autonomous vehicle is traveling on a highway, sensors detect a sudden decrease in the distance to the vehicle ahead. This data is pre-processed, and an AI model evaluates the risk in real time. If the risk is deemed high, an alert is generated. The alert is immediately displayed on a smartphone or head-mounted display, prompting the driver to apply emergency braking.

[1398] Prompt statement:

[1399] Write code to assess risk and generate an alert if the risk is high, based on the following autonomous vehicle sensor data:

[1400] Speed: float

[1401] Proximity: float

[1402] Acceleration: float

[1403] Include data pre-processing, training advanced AI models, data analysis, and alert generation as needed.

[1404] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1405] Step 1:

[1406] Data collection

[1407] The server collects real-time data from the autonomous vehicle's sensors and cameras, including speed, proximity, acceleration, etc. The collected data serves as input for subsequent processing steps.

[1408] Step 2:

[1409] Data Preprocessing

[1410] The server preprocesses the data collected in step 1. This preprocessing involves normalizing the data and performing feature engineering. For example, it normalizes the speed data to fit within a specified range and removes outliers. This preprocessed data serves as input for training and evaluating the AI ​​model.

[1411] Step 3:

[1412] Training an AI model

[1413] The server trains an AI model based on the preprocessed data from step 2. It uses algorithms such as random forest and deep learning to build a risk detection model. This trained model generates output results for detecting risks in real time.

[1414] Step 4:

[1415] Real-time risk detection

[1416] The server uses a trained AI model to analyze the data collected in real time and detect risks. For example, a sudden decrease in the distance ahead of the vehicle is considered a high risk. The results of this analysis serve as input data for subsequent alert generation.

[1417] Step 5:

[1418] Risk Scoring and Alerting

[1419] The server calculates a risk score and evaluates the severity based on the risks detected in step 4. If the risk score exceeds a certain threshold, it generates a risk alert and notifies the responsible party. The alert includes details of the risk, the scope of impact, and recommended countermeasures.

[1420] Step 6:

[1421] Alert Notification and Display

[1422] The device receives risk alerts sent from the server and notifies the user. The alerts are displayed via a user interface such as a smartphone or head-mounted display. Recommendations for risk countermeasures are also displayed on the device in real time.

[1423] Step 7:

[1424] Implementing risk response measures

[1425] The user checks the risk alerts and recommended countermeasures displayed on the device and takes specific countermeasures against the risk, such as applying emergency braking or proposing a route change. The results of this action become input data for the next feedback step.

[1426] Step 8:

[1427] Providing Feedback

[1428] Users can provide feedback on the results of their risk countermeasures to the server, which will be used to retrain the AI ​​model and improve the accuracy of the risk detection model.

[1429] 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.

[1430] The present invention relates to a system for companies to perform efficient and sophisticated risk management. In particular, by combining it with an emotion engine that recognizes user emotions, it provides a more accurate risk detection and alert system.

[1431] Overall flow

[1432] This system operates as follows between the server, terminal, and user: The server collects and analyzes data, trains AI models, detects risks, and generates alerts. The terminal is responsible for notifying alerts and displaying risk information, and recognizes user emotions using an emotion engine. The user implements risk countermeasures and provides feedback.

[1433] 1. Data Collection and Analysis

[1434] Server Action:

[1435] 1. Data Collection: The server periodically collects risk-related information from internal systems and external data sources, and stores and updates the collected data in a database.

[1436] 2. Data Preprocessing

[1437] Server Action:

[1438] 1. Data preprocessing: The server cleans the collected data, fills in missing data, removes outliers, and then standardizes the data and converts it into a format suitable for training AI models.

[1439] 3. Training the AI ​​model

[1440] Server Action:

[1441] 1. Model training: The server uses the preprocessed data to train an AI model for risk detection, applying machine learning algorithms to learn patterns that identify risks.

[1442] 2. Model evaluation: After training, the server evaluates the accuracy of the model and retrains it if necessary.

[1443] 4. Risk detection and alert generation

[1444] Server Action:

[1445] 1. Real-time data analysis: The server uses AI models to detect risks as business data flows in in real time.

[1446] 2. Risk Scoring: The server calculates a risk score for the detected risks and evaluates their severity.

[1447] 3. Alert Generation: If the risk score exceeds a pre-defined threshold, the server generates a risk alert and notifies the appropriate personnel.

[1448] 5. Alert Notification and Display

[1449] Terminal handling:

[1450] 1. Receiving alert: The terminal receives the risk alert sent from the server.

[1451] 2. Alert display: The device displays detailed information about the risk alert (risk details, scope of impact, recommended countermeasures) to the user.

[1452] 6. Emotion recognition

[1453] Terminal handling:

[1454] 1. Acquiring emotion data: Using the camera and microphone installed on the device, emotions are detected from the user's facial expressions and voice.

[1455] 2. Emotion Analysis: The emotion engine analyzes the emotion data to identify the user's emotional state, for example, to identify the level of stress or anxiety.

[1456] 7. Alert Tuning

[1457] Terminal handling:

[1458] 1. Emotion-based adjustment: The emotion engine adjusts the way risk alerts and recommended actions are displayed based on the user's emotions. For example, if the user is under a lot of stress, more concise and quick actions will be displayed first.

[1459] 8. Implementing risk response measures

[1460] User Action:

[1461] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[1462] 2. Implementing countermeasures: The user implements the necessary risk countermeasures based on the recommended countermeasures displayed, such as suspending transactions, reviewing processes, or launching an internal investigation.

[1463] 9. Providing Feedback

[1464] User Action:

[1465] 1. Providing feedback: Users provide feedback on the results of their risk response measures to the server. The feedback information is used to retrain the AI ​​model, contributing to improving its accuracy.

[1466] 2. Emotional feedback: Providing feedback, including emotional data, on how the user reacted to the alert.

[1467] Specific examples

[1468] As an example, consider a scenario where Company B is managing the risks of a new project. Here are some situations where the emotion engine identifies risks during the project and requires action:

[1469] 1. Data collection: The server collects work data and communication data of project participants.

[1470] 2. Training the AI ​​model: Train the risk detection model based on the pre-processed data.

[1471] 3. Risk detection: The server monitors the progress of the project and calculates a risk score when a risk is detected.

[1472] 4. Alert Generation: The server generates alerts for high-scoring risks and notifies relevant parties.

[1473] 5. Emotion recognition and adjustment: The device detects the project manager's emotions and, if stress levels are high, displays a brief alert to prioritize immediate response.

[1474] 6. Implement risk response measures: The user implements the recommended response measures and feeds the results back to the server.

[1475] 7. Emotional feedback: Feedback containing emotional data is sent to the server and used to improve next risk countermeasures.

[1476] In this way, the present invention enables companies to manage the complex risks they face in real time and utilizes sentiment data to enable more flexible and effective responses.

[1477] The processing flow will be explained below.

[1478] Step 1: Data collection and updates

[1479] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external data sources (e.g., legal databases, SNS). The collected data is stored and updated in a database.

[1480] Step 2: Data Preprocessing

[1481] The server cleans the collected data, fills in missing data, removes outliers, and standardizes (normalizes) the data, converting it into a format suitable for learning AI models, thereby improving the accuracy of model training.

[1482] Step 3: Training the AI ​​model

[1483] The server uses the preprocessed data to train the AI ​​model. Specifically, it uses machine learning algorithms (e.g., random forests, deep learning) to learn patterns to identify risks. After training, the server evaluates the accuracy of the model and retrains it if necessary.

[1484] Step 4: Real-time data analysis

[1485] The server uses AI models to detect risks as business data flows in in real time, using data stream analysis techniques to identify outliers and unexpected patterns.

[1486] Step 5: Risk Scoring

[1487] The server calculates a risk score for each detected risk, based on the scope of impact and probability of occurrence, and serves as a basis for assessing the severity of the risk.

[1488] Step 6: Alert Generation

[1489] The server generates risk alerts when the risk score exceeds a pre-defined threshold, with detailed information about the risk, its scope, and recommended actions to take.

[1490] Step 7: Alert Notifications

[1491] The terminal receives risk alerts sent from the server, and when an alert arrives, it notifies the user so that they can check it immediately.

[1492] Step 8: Displaying alerts

[1493] The device displays detailed information about the risk alert to the user, including the type of risk, the scope of impact, and specific recommended actions to take.

[1494] Step 9: Obtaining Emotion Data

[1495] The device is equipped with a camera and microphone, which are used to capture emotion data from the user's facial expressions and voice. The emotion engine analyzes this data and identifies the user's emotional state.

[1496] Step 10: Adjust alerts based on emotion

[1497] The device adjusts the way risk alerts and recommended actions are displayed based on the user's emotions as recognized by the emotion engine. For example, if the user is under a high level of stress, the device will prioritize displaying concise and specific actions.

[1498] Step 11: Implement risk responses

[1499] Users can check the alerts displayed on their devices and take necessary risk management measures based on the recommended actions presented, such as suspending trading, reviewing processes, or launching an internal investigation.

[1500] Step 12: Provide feedback

[1501] Users then provide feedback on the results of their risk response measures to the server. The feedback information is used to retrain the AI ​​model, helping to improve its accuracy. The feedback also includes user emotional data, which is used to improve future risk detection and response measures.

[1502] Example 2

[1503] 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."

[1504] Conventional risk management systems can detect risks and generate alerts, but they are unable to consider the user's emotional state, making it difficult to present accurate and prompt countermeasures. Furthermore, retraining including user emotional data to improve the accuracy of feedback has not been performed, limiting the improvement of the system's accuracy.

[1505] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1506] In this invention, the server includes means for collecting risk-related information from a data source, means for learning and building a risk detection model based on the collected information, means for detecting risks in real time using the learned risk detection model, means for generating a risk alert based on the detected risk and notifying a responsible person, means for displaying the generated risk alert and recommended countermeasures on a user terminal, means for acquiring and analyzing user emotion data, means for adjusting the display method of the risk alert and recommended countermeasures based on the emotion analysis results, and means for providing feedback on the results of the risk countermeasures implemented based on the risk alert. This enables accurate and prompt presentation of risk countermeasures that take the user's emotional state into consideration, and enables continuous improvement of the accuracy of the system using feedback.

[1507] "Data Sources" are internal and external information sources used to provide risk-related information.

[1508] A "risk detection model" is an AI model that is trained and constructed to identify and evaluate risks based on collected information.

[1509] "Means for detecting risks in real time" refers to a function that uses AI models to instantly identify risks when business data flows in.

[1510] A "Risk Alert" is a notification of a detected risk, including detailed information about the risk and recommended actions to take.

[1511] "Means for notifying the person in charge" refers to a communication means for notifying the person in charge of detected risk information, and includes, for example, email and push notification.

[1512] "User device" means a device used to display risk alerts and recommended countermeasures to a user, including a personal computer or smartphone.

[1513] "Emotion data" is data that indicates the emotional state of the user, obtained from facial expressions, tone of voice, and the like.

[1514] "Emotion analysis" is the process of identifying a user's emotional state based on acquired emotion data.

[1515] The "means for adjusting the display method" is a function for changing the display format of risk alerts and recommended countermeasures based on the results of sentiment analysis.

[1516] A "risk response" is a specific action or procedure taken based on a risk alert to reduce or avoid a risk.

[1517] "Feedback" is the process by which the user sends back to the server the results of the risk response actions they have taken and their emotional state.

[1518] This invention relates to a system for companies to perform efficient and sophisticated risk management. In particular, by combining an emotion engine that recognizes user emotions, it provides a more accurate risk detection and alert system. This system operates between a server, terminals, and users.

[1519] Server Processing

[1520] Data collection and analysis

[1521] The server periodically collects risk-related information from internal systems and external data sources (e.g., the company's ERP system, CRM database, or news feeds and social media). This information is stored in an integrated database and updated on demand for analysis. For example, the server retrieves economic data using APIs every night and imports it into the database.

[1522] Data Preprocessing

[1523] The server cleans the collected data, complements incomplete data, and removes outliers. Next, it standardizes the data and converts it into a format suitable for AI model training. For example, it automatically removes outliers such as "0" and "NULL" from the collected data and inserts average values ​​into missing values.

[1524] Training and evaluating AI models

[1525] The server uses the preprocessed data to train an AI model for risk detection. Specifically, it applies a machine learning algorithm (e.g., random forest) to learn patterns for identifying risks. For example, the server uses the training data to build a random forest model and extracts features of risk factors. After training, the server evaluates the accuracy of the model using test data and retrains it if the accuracy is insufficient. For example, the server evaluates the model's accuracy using cross-validation and retrains the model if the F1 score falls below 80%.

[1526] Risk detection and alert generation

[1527] The server analyzes business data in real time and detects risks using AI models. For example, the server receives new trading data every minute and immediately analyzes it using AI models. It calculates a risk score for the detected risks and evaluates their severity. For example, if the server detects abnormal trading activity, it assigns a risk score to that activity and determines the severity of the risk based on the score. If the risk score exceeds a set threshold, it generates a risk alert and sends a notification to the appropriate person in charge. For example, if the risk score exceeds 80, the server generates a risk alert and sends a notification to the email address of the designated person in charge.

[1528] Terminal handling

[1529] Alert Notification and Display

[1530] The device receives the risk alert sent from the server. For example, the device receives a push notification from the server and the alert is ready to be displayed. Then, the device displays detailed information about the risk alert (risk details, impact scope, and recommended countermeasures) to the user. For example, the device opens a pop-up window and notifies the user, "A new risk has been detected. Please check the impact scope and recommended countermeasures."

[1531] emotion recognition

[1532] The device uses a camera and microphone to acquire emotions from the user's facial expressions and tone of voice. For example, the device uses a camera to capture the user's facial expressions while checking an alert, and records the user's tone of voice with a microphone. The emotion engine analyzes the acquired emotion data and identifies the user's emotional state. For example, the emotion engine uses facial recognition technology to analyze the user's facial expressions and determine that the user is in a "high stress state."

[1533] Alert Tuning

[1534] The emotion engine recognizes the user's emotions and adjusts the way risk alerts and recommended solutions are displayed. For example, if the user is under a lot of stress, the device will prioritize displaying "simple and quick solutions" and postpone detailed explanations.

[1535] User Action

[1536] Implementing risk response measures

[1537] The user checks the details of the risk alert displayed on the terminal and takes necessary risk countermeasures based on the recommended countermeasures displayed. For example, the user reads the details of the alert displayed on the terminal screen, clicks the "pause trading" button, and sends instructions to the system.

[1538] Providing Feedback

[1539] The user feeds back the results of the risk countermeasures they have implemented to the server. The feedback information is used to retrain the AI ​​model. For example, the user inputs whether the implemented countermeasures were successful or not and sends that data to the server. Feedback including the user's emotional data is also provided and used for the next risk countermeasure. For example, the user inputs their emotional state after implementing a countermeasure into their device, and that information is sent to the server and used to improve the AI ​​model.

[1540] Examples and prompts

[1541] For example, consider a scenario where a company is starting a new project and managing the risks associated with it. Here's how the emotion engine might identify risks during the project and prompt a rapid response:

[1542] 1. Data Collection:

[1543] The server collects work and communication data from the project team, for example, it uses APIs to extract data from a project management tool.

[1544] 2. Training the AI ​​model:

[1545] Based on the preprocessed data, a risk detection model is trained using a framework such as TensorFlow. For example, the server uses project data to learn patterns of risk occurrence.

[1546] 3. Risk detection:

[1547] The server monitors the project management system (PMS) in real time, identifying and scoring risks. For example, the server periodically updates data and calculates risk scores in real time.

[1548] 4. Alert Generation:

[1549] If the risk score is high, the server sends a risk alert to the project team's chat tool (e.g., a team communication tool). For example, if a serious risk is detected, the server notifies the team that an "urgent risk has occurred."

[1550] 5. Emotion recognition and regulation:

[1551] The device detects the project manager's emotions and, if stress levels are high, the device will simplify the alert display and prioritize immediate action. For example, the device will display a "summary of countermeasures" and encourage the manager to check the details later.

[1552] 6. Implementing risk response measures:

[1553] The project manager implements the recommended action and sends the results to the server, for example, the project manager clicks the "Stop Transaction" button and records the results.

[1554] 7. Emotional Feedback:

[1555] The project manager reports their emotional state after implementing the countermeasures, and this information is used to plan the next risk countermeasure. For example, the project manager inputs their impression of "stress reduction," and the data is sent to the server.

[1556] Examples of prompt statements

[1557] Below are some example prompts to input to a generative AI model:

[1558] 1. Example prompt 1:

[1559] "Identify risk factors in new projects. Generate a risk score by predicting the likelihood of risk occurrence based on project team work data and historical risk data."

[1560] 2. Example prompt 2:

[1561] "Analyze the emotional state of the project leader and, if the stress level is high, adjust the alert display and suggest quick countermeasures. Capture emotional data using the camera and microphone and analyze it with the emotion engine."

[1562] In this way, the present invention manages the complex risks faced by companies in real time and enables more flexible and effective responses by utilizing user emotional data.

[1563] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1564] System program processing flow and explanation

[1565] Step 1: Data collection

[1566] Server Processing

[1567] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM databases) and external data sources (e.g., news feeds, economic data, social media) using APIs and scraping techniques.

[1568] Input: API call, scraping script

[1569] Output: Risk-related information dataset

[1570] Specific operation: For example, every night at night, the server retrieves economic data using an API and imports it into the database.

[1571] Step 2: Data Preprocessing

[1572] Server Processing

[1573] The server cleans the collected data, fills in incomplete data, removes outliers, and then standardizes the data and converts it into a format suitable for AI model training.

[1574] Input: Risk-related information dataset

[1575] Output: Preprocessed dataset

[1576] Specific operation: For example, automatically remove outliers such as "0" and "NULL" from collected data and insert the average value into missing values.

[1577] Step 3: Training the AI ​​model

[1578] Server Processing

[1579] The server uses the preprocessed data to train an AI model for risk detection, specifically by applying machine learning algorithms (e.g., random forests) to learn patterns to identify risks.

[1580] Input: Preprocessed dataset

[1581] Output: A trained AI model

[1582] Specific operations: For example, the server uses training data to build a random forest model and extract features of risk factors.

[1583] Step 4: Model evaluation

[1584] Server Processing

[1585] After training, the server evaluates the accuracy of the model using test data and retrains it if the accuracy is insufficient.

[1586] Input: trained AI model, test dataset

[1587] Output: Evaluation results, improved AI model if necessary

[1588] Specific behavior: For example, the server evaluates the accuracy of the model using cross-validation and retrains the model if the F1 score falls below 80%.

[1589] Step 5: Real-time data analysis

[1590] Server Processing

[1591] The server analyzes business data in real time and uses AI models to detect risks.

[1592] Input: Real-time business data

[1593] Output: Detected risk information

[1594] Specific operation: For example, the server receives new transaction data every minute and immediately analyzes it using an AI model.

[1595] Step 6: Risk Scoring

[1596] Server Processing

[1597] The server calculates a risk score for the detected risk and evaluates its severity.

[1598] Input: Detected risk information

[1599] Output: Risk score, risk severity rating

[1600] Specific operation: For example, if the server detects abnormal trading activity, it assigns a risk score to the activity and determines the severity of the risk based on the score.

[1601] Step 7: Alert Generation

[1602] Server Processing

[1603] If the risk score exceeds a set threshold, the server generates a risk alert and sends a notification to the appropriate personnel.

[1604] Inputs: Risk score, risk severity rating

[1605] Output: Risk alert notification

[1606] Specific behavior: For example, if the risk score exceeds 80, the server generates a risk alert and sends a notification to the email address of the designated person in charge.

[1607] Step 8: Receive and view alerts

[1608] Terminal handling

[1609] The terminal receives risk alerts sent from the server and displays detailed information about the risk alert (risk details, scope of impact, recommended countermeasures) to the user.

[1610] Input: Risk Alert Notification

[1611] Output: Alert details display

[1612] Specific behavior: For example, the device will open a pop-up window and notify you that "A new risk has been detected. Please check the scope of impact and recommended countermeasures."

[1613] Step 9: Acquire and analyze emotion data

[1614] Terminal handling

[1615] Using the device's built-in camera and microphone, emotions are acquired from the user's facial expressions and tone of voice. The emotion engine analyzes the acquired emotional data and identifies the user's emotional state.

[1616] Input: Camera video, microphone audio

[1617] Output: Emotion analysis results

[1618] Specific operation: For example, the device captures the user's facial expression while checking the alert with a camera and records the tone of voice with a microphone. The emotion engine uses facial recognition technology to analyze the user's facial expression and determine whether they are in a "high stress state."

[1619] Step 10: Adjust alerts based on emotion

[1620] Terminal handling

[1621] Tailor how risk alerts and recommended actions are presented based on the user's emotions as recognized by the emotion engine.

[1622] Input: Sentiment analysis results, risk alerts

[1623] Output: Adjusted alert display

[1624] Specific operation: For example, if the user is under a lot of stress, the device will prioritize displaying "simple and quick solutions" and postpone detailed explanations.

[1625] Step 11: Implement risk responses

[1626] User Action

[1627] The user checks the content of the risk alert displayed on the device and takes the necessary risk countermeasures based on the recommended countermeasures displayed.

[1628] Input: Risk alert, recommended action

[1629] Output: Risk response measures taken, response results

[1630] Specific actions: For example, the user reads the details of the alert displayed on the terminal screen, clicks the "Pause trading" button, and sends instructions to the system.

[1631] Step 12: Provide feedback and retrain the model

[1632] User and Server Processing

[1633] Users provide feedback on the results of their risk countermeasures to the server, which uses the feedback information to retrain the AI ​​model.

[1634] Input: Countermeasure execution results, emotional feedback

[1635] Output: Improved AI model

[1636] Specific operation: For example, the user inputs whether the implemented countermeasure was successful and sends the data to the server. The user also reports their emotional state after the countermeasure was implemented, and this information is sent to the server and used to improve the AI ​​model.

[1637] (Application example 2)

[1638] 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."

[1639] While conventional risk management systems are capable of detecting risks based on data collection and analysis, they lack the flexibility to respond flexibly, taking into account the user's emotional state. This can result in inappropriate adjustment of risk alerts and limited effectiveness in implementing countermeasures. Furthermore, the lack of user emotional feedback leads to insufficient improvement in the accuracy of risk management.

[1640] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1641] In this invention, the server includes: means for collecting risk-related information from a data source; means for learning and building a risk detection model based on the collected information; means for detecting risks in real time using the learned risk detection model; means for generating a risk alert based on the detected risk and notifying a responsible party; means for displaying the generated risk alert and recommended countermeasures on a user terminal; means for recognizing a user's emotions in real time using a camera and microphone mounted on the terminal; means for adjusting the display method of the risk alert based on the emotions recognized by the emotion engine; and means for feeding back the results of the risk countermeasures implemented based on the risk alert. This allows user emotion data to be incorporated into the risk management system, enabling more accurate risk detection and flexible responses.

[1642] "Data Source" means a source of information for collecting risk-related information, including internal systems and external databases.

[1643] A "risk detection model" is an AI model that is trained using machine learning algorithms to identify and assess risks based on collected information.

[1644] "Real-time" refers to the fact that information is processed and analyzed instantly, without any time lag.

[1645] A "risk alert" is a warning notification generated based on a detected risk, which includes details of the risk and recommended countermeasures.

[1646] "User terminal" refers to a device for displaying risk alerts and recommended countermeasures, including smartphones and smart glasses.

[1647] The "emotion engine" is software that uses the device's built-in camera and microphone to recognize and analyze the user's emotional state in real time.

[1648] "Risk score" is a numerical representation of the severity of the risk assessed by the AI ​​model.

[1649] "Feedback" refers to the results of risk response measures taken by the user and emotional data being sent back to the server, and is information used to retrain the AI ​​model.

[1650] "Preprocessing" is the process of cleaning collected data, filling in missing data, removing outliers, standardizing it, and converting it into a format suitable for learning AI models.

[1651] This invention relates to a smart logistics risk management system that improves work efficiency and safety within logistics centers. Specifically, it has the function of recognizing the emotional state of workers in real time through terminals equipped with an emotion engine and adjusting risk alerts. This system operates as follows between the server, terminals, and users.

[1652] Server Processing

[1653] The server handles data collection, data preprocessing, AI model training and retraining, risk detection, risk scoring, and alert generation. Specific software used includes machine learning algorithms and sentiment engines. Data is standardized, and risk detection models are trained using algorithms such as RandomForestClassifier.

[1654] Data collection and preprocessing

[1655] The server periodically collects risk-related information from internal systems and external databases, stores it in a database, and then pre-processes the data by cleaning it, removing outliers, and filling in missing data, converting it into a format suitable for training AI models.

[1656] Training and evaluating AI models

[1657] The risk detection model is trained based on the pre-processed data. After training, the model undergoes accuracy evaluation and is retrained as necessary. Emotion recognition data from the emotion engine is also integrated into the risk detection model, contributing to improving the model's accuracy.

[1658] Terminal handling

[1659] The device will recognize emotions in real time, display and adjust risk alerts, and collect user feedback. The hardware used includes smart glasses and smartphones with an integrated emotion engine.

[1660] Emotion Recognition and Alert Adjustment

[1661] Using cameras and microphones mounted on smart glasses or smartphones, the system detects emotions from the worker's facial expressions and voice. An emotion engine analyzes the emotional data and identifies the user's emotional state. Based on this information, the system adjusts the way risk alerts are displayed. For example, if stress levels are high, quick and concise alerts will be prioritized.

[1662] User Action

[1663] The user checks the alerts displayed on their device, takes recommended risk countermeasures, and then provides the results and emotional feedback to the server, which then retrains the AI ​​model and improves its accuracy.

[1664] Specific examples

[1665] Consider the following scenario as an example for Logistics Center A. The server collects work data and environmental data from cameras and sensors within the center, as well as facial expression and voice data from workers' smart glasses. Based on the pre-processed data, a risk detection model is trained and performs real-time data analysis to detect risks. A risk score indicates the severity, and an appropriate risk alert is generated and notified to the terminal. The terminal recognizes the user's emotional state in real time and adjusts the alert display method. Workers implement the displayed risk countermeasures and provide the results and emotional feedback to the server.

[1666] For example, the following prompt sentences are applied as example input to a generative AI model:

[1667] "A logistics center worker shows signs of high stress while performing tasks. Their facial expression indicates stress with a confidence level of 0.7. The current risk score for the ongoing task is 0.8. Provide a concise alert message."

[1668] This approach streamlines risk management within the distribution center and improves worker safety.

[1669] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1670] Step 1: Data collection

[1671] The server periodically collects risk-related information from internal systems and external databases and stores it in a database. Specific inputs include work data, environmental data, and communication data. This ensures the data necessary for training risk detection models and real-time analysis. The output is organized raw data stored in the database.

[1672] Step 2: Data Preprocessing

[1673] The server cleans the collected data, fills in missing data, and removes outliers. It then standardizes the data and converts it into a format suitable for AI model training. Specific operations include data standardization, outlier elimination, and missing value imputation. The input is raw data, and the output is preprocessed data.

[1674] Step 3: Training the AI ​​model

[1675] The server uses the preprocessed data to train an AI model for risk detection. It applies a machine learning algorithm (e.g., RandomForestClassifier) ​​to learn patterns to identify risks. After training, it evaluates the accuracy of the model and retrains it if necessary. The input is the preprocessed data, and the output is a trained risk detection model.

[1676] Step 4: Obtaining emotion recognition data

[1677] The device uses the on-board camera and microphone to recognize emotions from the user's facial expressions and voice in real time. The emotion engine analyzes the emotion data and identifies the user's emotional state. Specific operations include face detection, voice analysis, and emotion classification. The input is raw data acquired by the camera and microphone, and the output is analyzed emotion data.

[1678] Step 5: Risk detection and alert generation

[1679] The server analyzes incoming data in real time and detects risks using a trained risk detection model. It then calculates a risk score and evaluates the severity. If the risk score exceeds a set threshold, it generates a risk alert and notifies the appropriate personnel. The input is real-time data, and the output is the generated risk alert.

[1680] Step 6: Adjust alerts based on emotion

[1681] The device adjusts the display of risk alerts and recommended countermeasures based on the user's emotions identified by the emotion engine. For example, if the user is in a state of high stress, it prioritizes displaying simpler and quicker countermeasures. The input is the analyzed emotion data and the generated risk alert, and the output is the adjusted risk alert display.

[1682] Step 7: Implement risk responses

[1683] The user checks the risk alert displayed on the terminal and implements the recommended countermeasures. Specific actions include changing work methods, pausing processes, contacting other personnel, etc. The input is the adjusted risk alert, and the output is feedback on the implementation of the countermeasures.

[1684] Step 8: Provide feedback

[1685] The user provides the server with the results of the risk countermeasures they have implemented and their emotional feedback. The server then retrains the AI ​​model based on this feedback data to improve the model's accuracy. The input is the user's feedback data, and the output is the retrained AI model.

[1686] For example, the following prompt sentences are applied as example input to a generative AI model:

[1687] "A logistics center worker shows signs of high stress while performing tasks. Their facial expression indicates stress with a confidence level of 0.7. The current risk score for the ongoing task is 0.8. Provide a concise alert message."

[1688] 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.

[1689] 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.

[1690] 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.

[1691] [Fourth embodiment]

[1692] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1693] 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.

[1694] 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).

[1695] 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.

[1696] 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.

[1697] 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).

[1698] 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.

[1699] 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.

[1700] 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.

[1701] 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.

[1702] 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.

[1703] 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.

[1704] 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."

[1705] The present invention relates to a system for enabling a company to efficiently manage risk. The specific processing content of the program of this system and an embodiment thereof will be described below with specific examples.

[1706] Overall flow

[1707] This system operates between the server, terminals, and users as follows: The server collects and analyzes data, trains AI models, and detects risks and generates alerts. The terminals are responsible for displaying alert notifications and risk information, while users implement risk countermeasures and provide feedback.

[1708] 1. Data Collection and Analysis

[1709] Server Action:

[1710] 1. Data collection: The server collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external legal databases and social networking sites.

[1711] 2. Data Update: The server saves and updates the collected data to the database used by date and time.

[1712] 2. Learning and training the AI ​​model

[1713] Server Action:

[1714] 1. Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the AI ​​model.

[1715] 2. Model training: The server uses the pre-processed data to train an AI model for risk detection, leveraging machine learning algorithms (e.g., random forest, deep learning).

[1716] 3. Model evaluation: The server evaluates the accuracy of the model and retrains it if necessary.

[1717] 3. Risk detection and alert generation

[1718] Server Action:

[1719] 1. Real-time data analysis: The server analyzes incoming business data in real time and detects risks using AI models.

[1720] 2. Risk Scoring: The server calculates a risk score for each detected risk and evaluates its severity.

[1721] 3. Alert generation: If the risk score exceeds a certain threshold, the server generates an alert and notifies the appropriate personnel.

[1722] 4. Alert notification and display

[1723] Terminal handling:

[1724] 1. Receiving an alert: The terminal receives an alert sent from the server.

[1725] 2. Alert display: The terminal displays detailed information about the alert (risk details, scope of impact, recommended countermeasures) to the user.

[1726] 5. Risk Management

[1727] User Action:

[1728] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[1729] 2. Implementing countermeasures: The user implements the necessary risk countermeasures based on the recommended countermeasures presented, such as suspending transactions, reviewing processes, or initiating an internal investigation.

[1730] 3. Providing feedback: The user provides feedback on the results of the implementation to the server, and this information is used to retrain the model.

[1731] Specific examples

[1732] As an example, consider a scenario in which Company A detects reputation risks, including negative reactions on social media, when releasing a new product.

[1733] 1. Data Collection: The server collects social media data and internal marketing data.

[1734] 2. Training the AI ​​model: Train the risk detection model based on the pre-processed data.

[1735] 3. Risk detection: When a new product is released, the server detects negative feedback in real time and calculates a high risk score.

[1736] 4. Alert Generation: The server generates and notifies the marketing department personnel of a risk alert.

[1737] 5. Notification and response: The device receives an alert, the user checks the relevant information, and modifies the new marketing strategy based on the recommendations.

[1738] In this way, the present invention enables efficient management of the complex risks faced by companies and realizes prompt and appropriate responses.

[1739] The processing flow will be explained below.

[1740] Step 1: Data collection and updates

[1741] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external data sources (e.g., legal databases, SNS). The collected data is stored and updated in a database.

[1742] Step 2: Data Preprocessing

[1743] The server cleans the collected data, fills in missing data, removes outliers, and then standardizes (normalizes) the data and converts it into a format suitable for training AI models.

[1744] Step 3: Training the AI ​​model

[1745] The server uses the preprocessed data to train the AI ​​model, applying machine learning algorithms to learn patterns to identify risks. After training, the model's accuracy is evaluated and retrained if necessary.

[1746] Step 4: Real-time data analysis

[1747] The server uses AI models to detect risks as business data flows in in real time, using data stream analysis techniques to identify outliers and unexpected patterns.

[1748] Step 5: Risk Scoring

[1749] The server calculates a risk score for each detected risk, based on the scope of impact and probability of occurrence, and serves as a basis for assessing the severity of the risk.

[1750] Step 6: Alert Generation

[1751] The server generates risk alerts when the risk score exceeds a pre-defined threshold, with detailed information about the risk, its scope, and recommended actions to take.

[1752] Step 7: Alert Notifications

[1753] The terminal receives risk alerts sent from the server, and when an alert arrives, it notifies the user so that they can check it immediately.

[1754] Step 8: Displaying alerts

[1755] The device displays detailed information about the risk alert to the user, including the type of risk, the scope of impact, and specific recommended actions to take.

[1756] Step 9: Implement risk responses

[1757] Based on the displayed risk alerts and recommended actions, users can take necessary risk response measures, such as suspending trading, reviewing processes, or launching an internal investigation.

[1758] Step 10: Provide feedback

[1759] Users then provide feedback on the results of their risk response measures to the server, which uses the feedback information to subsequently retrain the AI ​​model, contributing to improving the model's accuracy.

[1760] Example 1

[1761] 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."

[1762] Companies are being asked to improve the efficiency of risk management and respond quickly, but many challenges exist in collecting information from diverse data sources, training AI models, detecting and notifying risks in real time, and incorporating feedback. In particular, the processes of data preprocessing and model retraining are complex, and integrating the entire system takes time and effort. Furthermore, real-time risk detection and accurate alert generation are difficult, which can lead to a lack of speed and accuracy in risk responses. It is desirable to resolve these issues.

[1763] 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.

[1764] In this invention, the server includes a means for collecting risk-related information from data sources, a means for storing and updating the collected information in a database by date and time, and a means for preprocessing the collected data and converting it into a format suitable for training a generative AI model, thereby enabling companies to efficiently manage risk using an integrated system.

[1765] "Data Source" is a collective term for internal systems and external sources used to collect risk-related information.

[1766] A "risk detection model" is a general term for algorithms and machine learning models that are learned and built based on collected data to detect risks.

[1767] "Real-time" refers to processing and analysis being carried out immediately, close to the moment data is generated.

[1768] "Risk alert" refers to a warning or caution message that is generated based on a detected risk and notified to a responsible person.

[1769] "User Terminal" refers collectively to any electronic device used to display risk alerts and recommended responses.

[1770] "Feedback" refers to providing information about the results of risk response measures to the server.

[1771] A "generative AI model" refers to a model built using machine learning algorithms based on collected data.

[1772] "Preprocessing" is the process of organizing and transforming collected data to convert it into a form suitable for training an AI model.

[1773] An "algorithm" refers to a combination of steps or computational methods for solving a specific problem.

[1774] A "hyperparameter" is a setting value during the training process of a machine learning model, and refers to an element that affects the performance of the model.

[1775] A "stream processing framework" refers to a software platform for processing and analyzing data in real time.

[1776] The present invention relates to a system for enabling a company to efficiently manage risk. The specific processing contents of the program of this system and an embodiment thereof will be described below.

[1777] Overall flow

[1778] This system operates between the server, terminals, and users as follows: The server collects and analyzes data, trains generative AI models, and detects risks and generates alerts. The terminals are responsible for notifying alerts and displaying risk information, while users implement risk countermeasures and provide feedback.

[1779] Data collection and analysis

[1780] Server Processing

[1781] 1. Data collection: The server obtains business data from internal systems (e.g., ERP systems, CRM systems) and also collects risk-related information from external legal databases and social media APIs.

[1782] The server obtains financial and supply chain data by calling the ERP system's API.

[1783] Use a social media API (e.g., Twitter API) to collect tweets related to a specific hashtag.

[1784] Use a web scraping tool (e.g., BeautifulSoup) to scrape new legal information from the legal database.

[1785] 2. Data update: The server saves and updates the collected data in a database (e.g., MySQL, PostgreSQL) according to the date and time.

[1786] The new data is stored in a temporary table and then reconciled with the existing database using a journal file to ensure data consistency.

[1787] If the data is updated, update the timestamp and insert the new data.

[1788] Learning and training AI models

[1789] Server Processing

[1790] 1. Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the generative AI model, specifically by imputing missing data, converting character data to numeric data, and normalizing the data.

[1791] Missing data will be imputed using the median.

[1792] Character data is converted to numeric data using one-hot encoding.

[1793] Normalization is performed to align the range of the data.

[1794] 2. Model training: The preprocessed data is used to train a generative AI model. Model training uses machine learning frameworks such as TensorFlow or scikit-learn, and employs algorithms such as random forests and deep learning.

[1795] The dataset is divided into batches for training.

[1796] Tune the hyperparameters of your model to get optimal performance.

[1797] 3. Model evaluation: Evaluate the accuracy of the trained model and retrain it if necessary.

[1798] For evaluation, we use a test dataset and calculate precision, recall, and F1 score.

[1799] If the accuracy is low, reset the hyperparameters and retrain.

[1800] Risk detection and alert generation

[1801] Server Processing

[1802] 1. Real-time data analysis: The server analyzes incoming business data in real time and uses generative AI models to detect risks, using stream processing frameworks such as Apache Kafka and Apache Flink.

[1803] Monitor real-time data streams and detect specific patterns.

[1804] Input data into the AI ​​model to obtain a predicted value.

[1805] 2. Risk Scoring: The server calculates a risk score for each detected risk and evaluates its severity.

[1806] Set scoring criteria according to the type of risk.

[1807] A score is calculated and evaluated based on the output value of the AI ​​model.

[1808] 3. Alert generation: If the risk score exceeds the set threshold, the server generates an alert and notifies the responsible person.

[1809] An email containing the alert content is created and sent using the SAP communication protocol.

[1810] Use a push notification service to send alerts to the mobile devices of personnel.

[1811] Alert Notification and Display

[1812] Terminal handling

[1813] 1. Receiving an alert: The device receives an alert sent from the server.

[1814] Receives push notifications from the notification server and displays them on the device.

[1815] 2. Alert display: Displays detailed information about the received alert (risk content, scope of impact, recommended countermeasures) to the user.

[1816] The mobile or desktop app will pop up an alert message.

[1817] Risk Management

[1818] User Action

[1819] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[1820] Check the alert displayed on the device screen in detail.

[1821] 2. Implementing Actions: The user implements the necessary risk response measures based on the recommended actions, which may include suspending trading, reviewing processes, or initiating an internal investigation.

[1822] Implement countermeasures based on company policy.

[1823] 3. Providing feedback: The results of the implemented countermeasures are provided to the server as feedback.

[1824] Feedback data is sent to the server using a dedicated input form.

[1825] Specific examples

[1826] Consider a scenario in which Company A detects reputation risks, including negative reactions on social media, when releasing a new product.

[1827] 1. Data Collection: The server collects social media data and internal marketing data.

[1828] 2. Training a generative AI model: Train a risk detection model based on the preprocessed data.

[1829] 3. Risk detection: When a new product is released, the server detects negative feedback in real time and calculates a high risk score.

[1830] 4. Alert generation: The server generates and notifies the marketing department personnel of a risk alert.

[1831] 5. Notification and response: The device receives an alert, the user checks the relevant information, and then a new marketing strategy is modified based on the recommendations.

[1832] Example prompt: "Describe a system that detects negative reactions on social media in real time when a company releases a new product and notifies the marketing department."

[1833] In this way, the present invention enables efficient management of the complex risks faced by companies and realizes prompt and appropriate responses.

[1834] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1835] Step 1: Data collection

[1836] The server retrieves business data from internal systems (e.g., ERP systems, CRM systems) and also collects risk-related information using external legal databases and SNS APIs.

[1837] Input: Business data, legal data, and social media data from data sources.

[1838] Specific behavior:

[1839] The server obtains financial and supply chain data by calling the ERP system's API.

[1840] Use a social media API (e.g., Twitter API) to collect tweets related to a specific hashtag.

[1841] Use a web scraping tool (e.g., BeautifulSoup) to scrape new legal information from the legal database.

[1842] Output: The retrieved data.

[1843] Step 2: Data Update

[1844] The server stores and updates the collected data in a database (e.g., MySQL, PostgreSQL) according to the date and time.

[1845] Input: The retrieved data.

[1846] Specific behavior:

[1847] Save the new data in a temporary table and compare it with the existing database to ensure data consistency during the update.

[1848] When data is updated, the timestamp is also updated and the latest data is saved.

[1849] Output: The updated database.

[1850] Step 3: Data Preprocessing

[1851] The server preprocesses the collected data and converts it into a format suitable for training the generative AI model, specifically by imputing missing data, converting character data to numeric data, and normalizing the data.

[1852] Input: Data from the updated database.

[1853] Specific behavior:

[1854] Missing data will be imputed using the median.

[1855] Character data is converted to numeric data using one-hot encoding.

[1856] Normalize the data to the range 0 to 1.

[1857] Output: Preprocessed data.

[1858] Step 4: Model training

[1859] The preprocessed data is used to train generative AI models, using machine learning frameworks including TensorFlow and scikit-learn.

[1860] Input: Preprocessed data.

[1861] Specific behavior:

[1862] Define a dataset and train batch by batch.

[1863] Set and tune the hyperparameters of your model for optimal performance.

[1864] Deep learning algorithms are used to train the models.

[1865] Output: A trained AI model.

[1866] Step 5: Model evaluation

[1867] Evaluate the accuracy of trained AI models and retrain them if necessary.

[1868] Input: A trained AI model, a test dataset.

[1869] Specific behavior:

[1870] The model is evaluated using a test dataset.

[1871] Calculate precision, recall, and F1 score to measure the performance of the model.

[1872] Adjust hyperparameters and retrain as needed.

[1873] Output: The evaluated AI model.

[1874] Step 6: Real-time data analysis

[1875] The server uses stream processing frameworks such as Apache Kafka and Apache Flink to analyze incoming business data in real time and detect risks using generative AI models.

[1876] Input: Business data flowing in in real time.

[1877] Specific behavior:

[1878] Monitor real-time data streams and detect important patterns.

[1879] The data is analyzed using an AI model to obtain predicted values.

[1880] Output: Risk detection results.

[1881] Step 7: Risk Scoring

[1882] The server calculates a risk score for each detected risk and evaluates its severity.

[1883] Input: Risk findings.

[1884] Specific behavior:

[1885] Use pre-defined scoring criteria for each type of risk.

[1886] A risk score is calculated based on the output value of the AI ​​model.

[1887] Output: Risk scoring results.

[1888] Step 8: Alert Generation

[1889] If the risk score exceeds a set threshold, the server generates an alert and notifies the responsible person.

[1890] Input: Risk scoring results.

[1891] Specific behavior:

[1892] Create and send an email containing the alert content.

[1893] Send alerts to personnel's mobile devices via push notification services.

[1894] Output: The generated alert.

[1895] Step 9: Receiving and viewing alerts

[1896] The terminal receives the alert sent from the server and displays the detailed information to the user.

[1897] Input: The generated alert.

[1898] Specific behavior:

[1899] Receives push notifications from the notification server and displays them on the device.

[1900] Pop up an alert message in the mobile or desktop app.

[1901] Output: The displayed alert.

[1902] Step 10: Risk response

[1903] The user checks the contents of the alert displayed on the terminal and implements the necessary risk countermeasures based on the recommended countermeasures.

[1904] Input: The displayed alert.

[1905] Specific behavior:

[1906] Check the alert displayed on the device screen.

[1907] In accordance with our internal policies, we will take appropriate action, which may include suspending business, reviewing processes, or initiating an internal investigation.

[1908] Output: The results of the implemented measures.

[1909] Step 11: Provide feedback

[1910] The user provides the results of the risk countermeasures implemented to the server as feedback.

[1911] Input: The result of the action taken.

[1912] Specific behavior:

[1913] Feedback data is sent to the server using a dedicated input form.

[1914] The server stores the feedback data and uses it to improve the model in the future.

[1915] Output: The newly collected feedback data.

[1916] (Application example 1)

[1917] 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."

[1918] With the evolution of autonomous vehicles, it has become increasingly important to adapt to real-time changes in traffic conditions and the surrounding environment and quickly and appropriately detect and manage risks. However, with conventional technology, it has been difficult to efficiently analyze the large amounts of data obtained from sensors and cameras and appropriately evaluate and notify risks. This has resulted in delayed responses to risks such as traffic accidents and breakdowns, which could result in serious damage. Therefore, improvements to real-time risk management systems are required.

[1919] 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.

[1920] In this invention, the server includes means for collecting risk-related information from a data source, means for learning and building a risk detection model based on the collected information, means for detecting risks in real time using the learned risk detection model, means for generating a risk alert based on the detected risk and notifying a responsible person, means for displaying the generated risk alert and recommended countermeasures on a user terminal, means for feeding back the results of executing the risk countermeasures based on the risk alert, means for collecting data in real time from sensors and cameras of the autonomous vehicle and analyzing traffic conditions and the surrounding environment, and means for displaying and executing countermeasures in the autonomous vehicle based on the risk alert, thereby enabling real-time risk management of the autonomous vehicle.

[1921] "Data sources" are sources of information used to collect risk-related information, such as internal systems, external legal databases, and social media.

[1922] A "risk detection model" is a trained AI model for detecting risks based on collected data, using machine learning algorithms such as random forest and deep learning.

[1923] A "risk alert" is a warning message that is generated based on a detected risk and is sent to a responsible person.

[1924] "User terminal" refers to a device that displays alert notifications and risk information. Specifically, this applies to smartphones and head-mounted displays.

[1925] A "sensor" is a device installed in an autonomous vehicle that measures speed, distance, acceleration, etc.

[1926] The "camera" is an image capture device that visualizes the surrounding environment of an autonomous vehicle and is used to grasp the situation in real time.

[1927] "Traffic conditions" refers to information that comprehensively indicates the position, speed, direction, etc. of vehicles and pedestrians on the road.

[1928] "Surrounding environment" refers to the physical and dynamic elements surrounding an autonomous vehicle, including buildings, other vehicles, and pedestrians.

[1929] "Feedback" is information used to evaluate the results of risk response measures taken and to help with the next model training.

[1930] "Real-time" refers to the near-instantaneous collection, analysis, notification, and response of information.

[1931] "Recommended countermeasures" refers to specific countermeasures or action plans suggested to users in response to detected risks.

[1932] The present invention relates to a system that collects and analyzes data in real time from sensors and cameras in autonomous vehicles, and quickly and appropriately detects and manages risks according to traffic conditions and the surrounding environment. Specific embodiments for carrying out the present invention are described below.

[1933] This system mainly consists of three components: a server, a terminal, and a user.

[1934] server

[1935] The server is responsible for data collection, analysis, training of AI models, risk detection, and alert generation. Specifically, it has the following functions:

[1936] Data collection: The server collects data in real time from the autonomous vehicle's sensors and cameras. The hardware used here includes the autonomous vehicle's speed sensor, distance sensor, acceleration sensor, camera, etc.

[1937] Data preprocessing: The server preprocesses the collected data and converts it into a format suitable for training the AI ​​model. Preprocessing includes data normalization and feature engineering. Software used includes Python, Pandas, and Numpy.

[1938] Training the AI ​​model: The server uses the preprocessed data to train a risk detection model. Machine learning algorithms include random forests and deep learning. Software used here includes scikit-learn.

[1939] Risk detection: Analyzes and detects risks in real time using trained AI models, and if a risk is detected, calculates a risk score and assesses its severity.

[1940] Alert generation: When a risk exceeds a certain threshold in importance, a risk alert is generated and the responsible person is notified.

[1941] Terminal

[1942] The device receives risk alerts from the server and notifies the user. Specifically, it has the following functions:

[1943] Receive alerts: Receive risk alerts sent from the server in real time.

[1944] Alert display: Received alerts are displayed visually and intuitively to the user on devices such as smartphones and head-mounted displays. These devices display detailed information about the alert (risk details, scope of impact, recommended countermeasures).

[1945] User

[1946] Users receive risk alerts via their devices and take appropriate countermeasures, which are detailed below:

[1947] Alert confirmation: The user confirms the content of the risk alert displayed on the device.

[1948] Countermeasure Execution: Based on the proposed countermeasures, implement the necessary risk countermeasures, such as applying emergency braking or proposing a route change.

[1949] Providing feedback: The results of the implemented countermeasures are fed back to the server. This information is used for the next retraining of the AI ​​model.

[1950] Examples and prompts

[1951] Examples:

[1952] When an autonomous vehicle is traveling on a highway, sensors detect a sudden decrease in the distance to the vehicle ahead. This data is pre-processed, and an AI model evaluates the risk in real time. If the risk is deemed high, an alert is generated. The alert is immediately displayed on a smartphone or head-mounted display, prompting the driver to apply emergency braking.

[1953] Prompt statement:

[1954] Write code to assess risk and generate an alert if the risk is high, based on the following autonomous vehicle sensor data:

[1955] Speed: float

[1956] Proximity: float

[1957] Acceleration: float

[1958] Include data pre-processing, training advanced AI models, data analysis, and alert generation as needed.

[1959] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1960] Step 1:

[1961] Data collection

[1962] The server collects real-time data from the autonomous vehicle's sensors and cameras, including speed, proximity, acceleration, etc. The collected data serves as input for subsequent processing steps.

[1963] Step 2:

[1964] Data Preprocessing

[1965] The server preprocesses the data collected in step 1. This preprocessing involves normalizing the data and performing feature engineering. For example, it normalizes the speed data to fit within a specified range and removes outliers. This preprocessed data serves as input for training and evaluating the AI ​​model.

[1966] Step 3:

[1967] Training an AI model

[1968] The server trains an AI model based on the preprocessed data from step 2. It uses algorithms such as random forest and deep learning to build a risk detection model. This trained model generates output results for detecting risks in real time.

[1969] Step 4:

[1970] Real-time risk detection

[1971] The server uses a trained AI model to analyze the data collected in real time and detect risks. For example, a sudden decrease in the distance ahead of the vehicle is considered a high risk. The results of this analysis serve as input data for subsequent alert generation.

[1972] Step 5:

[1973] Risk Scoring and Alerting

[1974] The server calculates a risk score and evaluates the severity based on the risks detected in step 4. If the risk score exceeds a certain threshold, it generates a risk alert and notifies the responsible party. The alert includes details of the risk, the scope of impact, and recommended countermeasures.

[1975] Step 6:

[1976] Alert Notification and Display

[1977] The device receives risk alerts sent from the server and notifies the user. The alerts are displayed via a user interface such as a smartphone or head-mounted display. Recommendations for risk countermeasures are also displayed on the device in real time.

[1978] Step 7:

[1979] Implementing risk response measures

[1980] The user checks the risk alerts and recommended countermeasures displayed on the device and takes specific countermeasures against the risk, such as applying emergency braking or proposing a route change. The results of this action become input data for the next feedback step.

[1981] Step 8:

[1982] Providing Feedback

[1983] Users can provide feedback on the results of their risk countermeasures to the server, which will be used to retrain the AI ​​model and improve the accuracy of the risk detection model.

[1984] 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.

[1985] The present invention relates to a system for companies to perform efficient and sophisticated risk management. In particular, by combining it with an emotion engine that recognizes user emotions, it provides a more accurate risk detection and alert system.

[1986] Overall flow

[1987] This system operates as follows between the server, terminal, and user: The server collects and analyzes data, trains AI models, detects risks, and generates alerts. The terminal is responsible for notifying alerts and displaying risk information, and recognizes user emotions using an emotion engine. The user implements risk countermeasures and provides feedback.

[1988] 1. Data Collection and Analysis

[1989] Server Action:

[1990] 1. Data Collection: The server periodically collects risk-related information from internal systems and external data sources, and stores and updates the collected data in a database.

[1991] 2. Data Preprocessing

[1992] Server Action:

[1993] 1. Data preprocessing: The server cleans the collected data, fills in missing data, removes outliers, and then standardizes the data and converts it into a format suitable for training AI models.

[1994] 3. Training the AI ​​model

[1995] Server Action:

[1996] 1. Model training: The server uses the preprocessed data to train an AI model for risk detection, applying machine learning algorithms to learn patterns that identify risks.

[1997] 2. Model evaluation: After training, the server evaluates the accuracy of the model and retrains it if necessary.

[1998] 4. Risk detection and alert generation

[1999] Server Action:

[2000] 1. Real-time data analysis: The server uses AI models to detect risks as business data flows in in real time.

[2001] 2. Risk Scoring: The server calculates a risk score for the detected risks and evaluates their severity.

[2002] 3. Alert Generation: If the risk score exceeds a pre-defined threshold, the server generates a risk alert and notifies the appropriate personnel.

[2003] 5. Alert Notification and Display

[2004] Terminal handling:

[2005] 1. Receiving alert: The terminal receives the risk alert sent from the server.

[2006] 2. Alert display: The device displays detailed information about the risk alert (risk details, scope of impact, recommended countermeasures) to the user.

[2007] 6. Emotion recognition

[2008] Terminal handling:

[2009] 1. Acquiring emotion data: Using the camera and microphone installed on the device, emotions are detected from the user's facial expressions and voice.

[2010] 2. Emotion Analysis: The emotion engine analyzes the emotion data to identify the user's emotional state, for example, to identify the level of stress or anxiety.

[2011] 7. Alert Tuning

[2012] Terminal handling:

[2013] 1. Emotion-based adjustment: The emotion engine adjusts the way risk alerts and recommended actions are displayed based on the user's emotions. For example, if the user is under a lot of stress, more concise and quick actions will be displayed first.

[2014] 8. Implementing risk response measures

[2015] User Action:

[2016] 1. Confirm the alert: The user confirms the content of the alert displayed on the terminal.

[2017] 2. Implementing countermeasures: The user implements the necessary risk countermeasures based on the recommended countermeasures displayed, such as suspending transactions, reviewing processes, or launching an internal investigation.

[2018] 9. Providing Feedback

[2019] User Action:

[2020] 1. Providing feedback: Users provide feedback on the results of their risk response measures to the server. The feedback information is used to retrain the AI ​​model, contributing to improving its accuracy.

[2021] 2. Emotional feedback: Providing feedback, including emotional data, on how the user reacted to the alert.

[2022] Specific examples

[2023] As an example, consider a scenario where Company B is managing the risks of a new project. Here are some situations where the emotion engine identifies risks during the project and requires action:

[2024] 1. Data collection: The server collects work data and communication data of project participants.

[2025] 2. Training the AI ​​model: Train the risk detection model based on the pre-processed data.

[2026] 3. Risk detection: The server monitors the progress of the project and calculates a risk score when a risk is detected.

[2027] 4. Alert Generation: The server generates alerts for high-scoring risks and notifies relevant parties.

[2028] 5. Emotion recognition and adjustment: The device detects the project manager's emotions and, if stress levels are high, displays a brief alert to prioritize immediate response.

[2029] 6. Implement risk response measures: The user implements the recommended response measures and feeds the results back to the server.

[2030] 7. Emotional feedback: Feedback containing emotional data is sent to the server and used to improve next risk countermeasures.

[2031] In this way, the present invention enables companies to manage the complex risks they face in real time and utilizes sentiment data to enable more flexible and effective responses.

[2032] The processing flow will be explained below.

[2033] Step 1: Data collection and updates

[2034] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM systems) and external data sources (e.g., legal databases, SNS). The collected data is stored and updated in a database.

[2035] Step 2: Data Preprocessing

[2036] The server cleans the collected data, fills in missing data, removes outliers, and standardizes (normalizes) the data, converting it into a format suitable for learning AI models, thereby improving the accuracy of model training.

[2037] Step 3: Training the AI ​​model

[2038] The server uses the preprocessed data to train the AI ​​model. Specifically, it uses machine learning algorithms (e.g., random forests, deep learning) to learn patterns to identify risks. After training, the server evaluates the accuracy of the model and retrains it if necessary.

[2039] Step 4: Real-time data analysis

[2040] The server uses AI models to detect risks as business data flows in in real time, using data stream analysis techniques to identify outliers and unexpected patterns.

[2041] Step 5: Risk Scoring

[2042] The server calculates a risk score for each detected risk, based on the scope of impact and probability of occurrence, and serves as a basis for assessing the severity of the risk.

[2043] Step 6: Alert Generation

[2044] The server generates risk alerts when the risk score exceeds a pre-defined threshold, with detailed information about the risk, its scope, and recommended actions to take.

[2045] Step 7: Alert Notifications

[2046] The terminal receives risk alerts sent from the server, and when an alert arrives, it notifies the user so that they can check it immediately.

[2047] Step 8: Displaying alerts

[2048] The device displays detailed information about the risk alert to the user, including the type of risk, the scope of impact, and specific recommended actions to take.

[2049] Step 9: Obtaining Emotion Data

[2050] The device is equipped with a camera and microphone, which are used to capture emotion data from the user's facial expressions and voice. The emotion engine analyzes this data and identifies the user's emotional state.

[2051] Step 10: Adjust alerts based on emotion

[2052] The device adjusts the way risk alerts and recommended actions are displayed based on the user's emotions as recognized by the emotion engine. For example, if the user is under a high level of stress, the device will prioritize displaying concise and specific actions.

[2053] Step 11: Implement risk responses

[2054] Users can check the alerts displayed on their devices and take necessary risk management measures based on the recommended actions presented, such as suspending trading, reviewing processes, or launching an internal investigation.

[2055] Step 12: Provide feedback

[2056] Users then provide feedback on the results of their risk response measures to the server. The feedback information is used to retrain the AI ​​model, helping to improve its accuracy. The feedback also includes user emotional data, which is used to improve future risk detection and response measures.

[2057] Example 2

[2058] 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."

[2059] Conventional risk management systems can detect risks and generate alerts, but they are unable to consider the user's emotional state, making it difficult to present accurate and prompt countermeasures. Furthermore, retraining including user emotional data to improve the accuracy of feedback has not been performed, limiting the improvement of the system's accuracy.

[2060] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[2061] In this invention, the server includes means for collecting risk-related information from a data source, means for learning and building a risk detection model based on the collected information, means for detecting risks in real time using the learned risk detection model, means for generating a risk alert based on the detected risk and notifying a responsible person, means for displaying the generated risk alert and recommended countermeasures on a user terminal, means for acquiring and analyzing user emotion data, means for adjusting the display method of the risk alert and recommended countermeasures based on the emotion analysis results, and means for providing feedback on the results of the risk countermeasures implemented based on the risk alert. This enables accurate and prompt presentation of risk countermeasures that take the user's emotional state into consideration, and enables continuous improvement of the accuracy of the system using feedback.

[2062] "Data Sources" are internal and external information sources used to provide risk-related information.

[2063] A "risk detection model" is an AI model that is trained and constructed to identify and evaluate risks based on collected information.

[2064] "Means for detecting risks in real time" refers to a function that uses AI models to instantly identify risks when business data flows in.

[2065] A "Risk Alert" is a notification of a detected risk, including detailed information about the risk and recommended actions to take.

[2066] "Means for notifying the person in charge" refers to a communication means for notifying the person in charge of detected risk information, and includes, for example, email and push notification.

[2067] "User device" means a device used to display risk alerts and recommended countermeasures to a user, including a personal computer or smartphone.

[2068] "Emotion data" is data that indicates the emotional state of the user, obtained from facial expressions, tone of voice, and the like.

[2069] "Emotion analysis" is the process of identifying a user's emotional state based on acquired emotion data.

[2070] The "means for adjusting the display method" is a function for changing the display format of risk alerts and recommended countermeasures based on the results of sentiment analysis.

[2071] A "risk response" is a specific action or procedure taken based on a risk alert to reduce or avoid a risk.

[2072] "Feedback" is the process by which the user sends back to the server the results of the risk response actions they have taken and their emotional state.

[2073] This invention relates to a system for companies to perform efficient and sophisticated risk management. In particular, by combining an emotion engine that recognizes user emotions, it provides a more accurate risk detection and alert system. This system operates between a server, terminals, and users.

[2074] Server Processing

[2075] Data collection and analysis

[2076] The server periodically collects risk-related information from internal systems and external data sources (e.g., the company's ERP system, CRM database, or news feeds and social media). This information is stored in an integrated database and updated on demand for analysis. For example, the server retrieves economic data using APIs every night and imports it into the database.

[2077] Data Preprocessing

[2078] The server cleans the collected data, complements incomplete data, and removes outliers. Next, it standardizes the data and converts it into a format suitable for AI model training. For example, it automatically removes outliers such as "0" and "NULL" from the collected data and inserts average values ​​into missing values.

[2079] Training and evaluating AI models

[2080] The server uses the preprocessed data to train an AI model for risk detection. Specifically, it applies a machine learning algorithm (e.g., random forest) to learn patterns for identifying risks. For example, the server uses the training data to build a random forest model and extracts features of risk factors. After training, the server evaluates the accuracy of the model using test data and retrains it if the accuracy is insufficient. For example, the server evaluates the model's accuracy using cross-validation and retrains the model if the F1 score falls below 80%.

[2081] Risk detection and alert generation

[2082] The server analyzes business data in real time and detects risks using AI models. For example, the server receives new trading data every minute and immediately analyzes it using AI models. It calculates a risk score for the detected risks and evaluates their severity. For example, if the server detects abnormal trading activity, it assigns a risk score to that activity and determines the severity of the risk based on the score. If the risk score exceeds a set threshold, it generates a risk alert and sends a notification to the appropriate person in charge. For example, if the risk score exceeds 80, the server generates a risk alert and sends a notification to the email address of the designated person in charge.

[2083] Terminal handling

[2084] Alert Notification and Display

[2085] The device receives the risk alert sent from the server. For example, the device receives a push notification from the server and the alert is ready to be displayed. Then, the device displays detailed information about the risk alert (risk details, impact scope, and recommended countermeasures) to the user. For example, the device opens a pop-up window and notifies the user, "A new risk has been detected. Please check the impact scope and recommended countermeasures."

[2086] emotion recognition

[2087] The device uses a camera and microphone to acquire emotions from the user's facial expressions and tone of voice. For example, the device uses a camera to capture the user's facial expressions while checking an alert, and records the user's tone of voice with a microphone. The emotion engine analyzes the acquired emotion data and identifies the user's emotional state. For example, the emotion engine uses facial recognition technology to analyze the user's facial expressions and determine that the user is in a "high stress state."

[2088] Alert Tuning

[2089] The emotion engine recognizes the user's emotions and adjusts the way risk alerts and recommended solutions are displayed. For example, if the user is under a lot of stress, the device will prioritize displaying "simple and quick solutions" and postpone detailed explanations.

[2090] User Action

[2091] Implementing risk response measures

[2092] The user checks the details of the risk alert displayed on the terminal and takes necessary risk countermeasures based on the recommended countermeasures displayed. For example, the user reads the details of the alert displayed on the terminal screen, clicks the "pause trading" button, and sends instructions to the system.

[2093] Providing Feedback

[2094] The user feeds back the results of the risk countermeasures they have implemented to the server. The feedback information is used to retrain the AI ​​model. For example, the user inputs whether the implemented countermeasures were successful or not and sends that data to the server. Feedback including the user's emotional data is also provided and used for the next risk countermeasure. For example, the user inputs their emotional state after implementing a countermeasure into their device, and that information is sent to the server and used to improve the AI ​​model.

[2095] Examples and prompts

[2096] For example, consider a scenario where a company is starting a new project and managing the risks associated with it. Here's how the emotion engine might identify risks during the project and prompt a rapid response:

[2097] 1. Data Collection:

[2098] The server collects work and communication data from the project team, for example, it uses APIs to extract data from a project management tool.

[2099] 2. Training the AI ​​model:

[2100] Based on the preprocessed data, a risk detection model is trained using a framework such as TensorFlow. For example, the server uses project data to learn patterns of risk occurrence.

[2101] 3. Risk detection:

[2102] The server monitors the project management system (PMS) in real time, identifying and scoring risks. For example, the server periodically updates data and calculates risk scores in real time.

[2103] 4. Alert Generation:

[2104] If the risk score is high, the server sends a risk alert to the project team's chat tool (e.g., a team communication tool). For example, if a serious risk is detected, the server notifies the team that an "urgent risk has occurred."

[2105] 5. Emotion recognition and regulation:

[2106] The device detects the project manager's emotions and, if stress levels are high, the device will simplify the alert display and prioritize immediate action. For example, the device will display a "summary of countermeasures" and encourage the manager to check the details later.

[2107] 6. Implementing risk response measures:

[2108] The project manager implements the recommended action and sends the results to the server, for example, the project manager clicks the "Stop Transaction" button and records the results.

[2109] 7. Emotional Feedback:

[2110] The project manager reports their emotional state after implementing the countermeasures, and this information is used to plan the next risk countermeasure. For example, the project manager inputs their impression of "stress reduction," and the data is sent to the server.

[2111] Examples of prompt statements

[2112] Below are some example prompts to input to a generative AI model:

[2113] 1. Example prompt 1:

[2114] "Identify risk factors in new projects. Generate a risk score by predicting the likelihood of risk occurrence based on project team work data and historical risk data."

[2115] 2. Example prompt 2:

[2116] "Analyze the emotional state of the project leader and, if the stress level is high, adjust the alert display and suggest quick countermeasures. Capture emotional data using the camera and microphone and analyze it with the emotion engine."

[2117] In this way, the present invention manages the complex risks faced by companies in real time and enables more flexible and effective responses by utilizing user emotional data.

[2118] The flow of the identification process in the second embodiment will be described with reference to FIG.

[2119] System program processing flow and explanation

[2120] Step 1: Data collection

[2121] Server Processing

[2122] The server periodically collects risk-related information from internal systems (e.g., ERP systems, CRM databases) and external data sources (e.g., news feeds, economic data, social media) using APIs and scraping techniques.

[2123] Input: API call, scraping script

[2124] Output: Risk-related information dataset

[2125] Specific operation: For example, every night at night, the server retrieves economic data using an API and imports it into the database.

[2126] Step 2: Data Preprocessing

[2127] Server Processing

[2128] The server cleans the collected data, fills in incomplete data, removes outliers, and then standardizes the data and converts it into a format suitable for AI model training.

[2129] Input: Risk-related information dataset

[2130] Output: Preprocessed dataset

[2131] Specific operation: For example, automatically remove outliers such as "0" and "NULL" from collected data and insert the average value into missing values.

[2132] Step 3: Training the AI ​​model

[2133] Server Processing

[2134] The server uses the preprocessed data to train an AI model for risk detection, specifically by applying machine learning algorithms (e.g., random forests) to learn patterns to identify risks.

[2135] Input: Preprocessed dataset

[2136] Output: A trained AI model

[2137] Specific operations: For example, the server uses training data to build a random forest model and extract features of risk factors.

[2138] Step 4: Model evaluation

[2139] Server Processing

[2140] After training, the server evaluates the accuracy of the model using test data and retrains it if the accuracy is insufficient.

[2141] Input: trained AI model, test dataset

[2142] Output: Evaluation results, improved AI model if necessary

[2143] Specific behavior: For example, the server evaluates the accuracy of the model using cross-validation and retrains the model if the F1 score falls below 80%.

[2144] Step 5: Real-time data analysis

[2145] Server Processing

[2146] The server analyzes business data in real time and uses AI models to detect risks.

[2147] Input: Real-time business data

[2148] Output: Detected risk information

[2149] Specific operation: For example, the server receives new transaction data every minute and immediately analyzes it using an AI model.

[2150] Step 6: Risk Scoring

[2151] Server Processing

[2152] The server calculates a risk score for the detected risk and evaluates its severity.

[2153] Input: Detected risk information

[2154] Output: Risk score, risk severity rating

[2155] Specific operation: For example, if the server detects abnormal trading activity, it assigns a risk score to the activity and determines the severity of the risk based on the score.

[2156] Step 7: Alert Generation

[2157] Server Processing

[2158] If the risk score exceeds a set threshold, the server generates a risk alert and sends a notification to the appropriate personnel.

[2159] Inputs: Risk score, risk severity rating

[2160] Output: Risk alert notification

[2161] Specific behavior: For example, if the risk score exceeds 80, the server generates a risk alert and sends a notification to the email address of the designated person in charge.

[2162] Step 8: Receive and view alerts

[2163] Terminal handling

[2164] The terminal receives risk alerts sent from the server and displays detailed information about the risk alert (risk details, scope of impact, recommended countermeasures) to the user.

[2165] Input: Risk Alert Notification

[2166] Output: Alert details display

[2167] Specific behavior: For example, the device will open a pop-up window and notify you that "A new risk has been detected. Please check the scope of impact and recommended countermeasures."

[2168] Step 9: Acquire and analyze emotion data

[2169] Terminal handling

[2170] Using the device's built-in camera and microphone, emotions are acquired from the user's facial expressions and tone of voice. The emotion engine analyzes the acquired emotional data and identifies the user's emotional state.

[2171] Input: Camera video, microphone audio

[2172] Output: Emotion analysis results

[2173] Specific operation: For example, the device captures the user's facial expression while checking the alert with a camera and records the tone of voice with a microphone. The emotion engine uses facial recognition technology to analyze the user's facial expression and determine whether they are in a "high stress state."

[2174] Step 10: Adjust alerts based on emotion

[2175] Terminal handling

[2176] Tailor how risk alerts and recommended actions are presented based on the user's emotions as recognized by the emotion engine.

[2177] Input: Sentiment analysis results, risk alerts

[2178] Output: Adjusted alert display

[2179] Specific operation: For example, if the user is under a lot of stress, the device will prioritize displaying "simple and quick solutions" and postpone detailed explanations.

[2180] Step 11: Implement risk responses

[2181] User Action

[2182] The user checks the content of the risk alert displayed on the device and takes the necessary risk countermeasures based on the recommended countermeasures displayed.

[2183] Input: Risk alert, recommended action

[2184] Output: Risk response measures taken, response results

[2185] Specific actions: For example, the user reads the details of the alert displayed on the terminal ...

Claims

1. a means of collecting risk-related information from data sources; and A means of learning and building risk detection models based on the collected information; A means for detecting risks in real time using the trained risk detection model; A means of generating risk alerts based on detected risks and notifying relevant personnel; a means for displaying the generated risk alert and recommended countermeasures on a user terminal; A means for providing feedback on the results of risk response measures based on risk alerts; A system including:

2. A means of preprocessing the collected data and converting it into a format suitable for training AI models; A means of training an AI model on the preprocessed data; and A means to evaluate the accuracy of trained AI models and retrain them if necessary; and The system of claim 1 further comprising:

3. A means to calculate risk scores and assess the severity of risks detected by the AI ​​model in real time; A means to include information in generated risk alerts, including risk details, scope of impact, and recommended remediation actions; A means to retrain and improve the AI ​​model based on user feedback data; and The system of claim 1 further comprising:

Images