Active anti-money laundering early warning method and system based on generative large language model

By adopting a proactive anti-money laundering early warning method based on a generative large language model, the problem of insufficient adaptability of bank anti-money laundering systems to new money laundering methods is solved, realizing fully automated and intelligent anti-money laundering early warning, and improving the real-time performance and accuracy of the system.

CN118982418APending Publication Date: 2024-11-19CHINA CITIC BANK CO LTD
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202411026693.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

Existing bank anti-money laundering systems lack independent logical reasoning capabilities, making it difficult to cope with new money laundering methods. This leads to the failure of pre-emptive prevention and in-process early warning mechanisms, and they are unable to meet the increasingly severe anti-money laundering needs.

Method used

An active anti-money laundering early warning method based on generative large language models is adopted. By establishing a set of prompt words, generating simulated transaction data, and training a rule scanning model, the logical reasoning and generalization capabilities of large language models are utilized to achieve automated and intelligent scanning and early warning of suspicious accounts and transactions.

Benefits of technology

It has achieved full automation and intelligence of the bank's anti-money laundering system, improved real-time performance and accuracy, effectively detected abnormal transaction behavior, and improved the execution efficiency of the anti-money laundering system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982418A_ABST
    Figure CN118982418A_ABST
Patent Text Reader

Abstract

The invention relates to an active anti-money laundering early warning method and system based on a generative large language model, and the method comprises the steps: generating a simulation transaction data set according to a prompt word set through a preset large language model, and enabling the simulation transaction data set to be used for executing model training operation, and executing suspicious pre-judgment operation on the real-time transaction data by using a preset big language model, and calling a rule scanning model to execute risk scanning operation on transaction items corresponding to the real-time transaction data. The functions of suspicious account scanning, suspicious transaction early warning, involved transaction scanning and the like are achieved, abnormal points in mass transactions are found according to natural language description for money laundering criminal behaviors, and compared with a traditional commercial bank anti-money laundering system, full-process automation and intelligentization of three stages of beforehand prevention, in-event early warning and after-event check are achieved, and the efficiency of money laundering is improved. And the real-time performance, the accuracy and the execution efficiency of the bank anti-money laundering system are greatly improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the fields of financial security, anti-money laundering risk identification and data processing technology, and in particular to an active anti-money laundering early warning method and system based on a generative large language model. Background Art

[0002] Anti-money laundering is an important part of bank risk prevention. In specific implementation, the anti-money laundering work of banks mainly includes: customer identification. When banks conduct business, they need to identify customers, verify customer identity information, and ensure the authenticity and integrity of customer information, and understand the true identity and background of customers; large and suspicious transaction reports. Large transactions refer to transactions exceeding a certain amount, while suspicious transactions refer to transactions that are inconsistent with the customer's identity, financial status or business. Once banks discover such transactions, they need to report them to relevant regulatory agencies immediately; risk assessment and classification management. Banks need to conduct risk assessments on customers and classify them according to their risk levels. For high-risk customers, banks need to strengthen monitoring and review to ensure that their transaction behaviors comply with laws, regulations and regulatory requirements. In addition, banks also need to establish a sound internal control mechanism to ensure that employees understand and comply with anti-money laundering laws and regulations, and regularly conduct anti-money laundering training for employees to improve their anti-money laundering awareness and capabilities.

[0003] At present, the anti-money laundering work of most banks involves three stages: pre-prevention, in-process warning, and post-event inspection. Among them, pre-prevention is mainly carried out through the inspection of suspicious accounts. Under the existing technology, the system will conduct risk assessment on accounts according to preset rules and models, and identify accounts that may have money laundering risks; in-process warning uses large-amount suspicious transaction warning systems and blacklist mechanisms to monitor ongoing transactions in real time. Once the transaction amount exceeds the preset threshold or is associated with an account on the blacklist, the system will immediately issue a warning signal; post-event inspection scans the details of transactions that have occurred, and uses data analysis technology to check whether there is any suspicious behavior, which usually includes a comprehensive analysis of transaction patterns, frequencies, amounts, etc. to discover abnormal transaction behaviors.

[0004] It can be seen that the existing anti-money laundering systems of banks based on traditional machine learning and deep learning algorithms generate prediction models by fitting the data between feature data and abnormal results. In addition, the existing anti-money laundering systems of banks can only detect whether there are abnormalities in transaction behaviors with similar rules to the training data, and do not have independent logical reasoning capabilities. This makes the existing anti-money laundering systems easy to be targeted by the black industry and evade anti-money laundering inspections, especially for newly emerging money laundering criminal methods. Often, only when multiple money laundering transactions have been completed and investigated by regulatory agencies, can the existing bank anti-money laundering systems timely adjust the relevant rules of the anti-money laundering system according to the notification of regulatory agencies. It is completely unable to continuously adjust to the changing criminal methods for new anti-money laundering methods, and there is a serious lag in response, resulting in the complete failure of the pre-prevention and in-process warning mechanisms, and it cannot meet the increasingly severe anti-money laundering work needs. Summary of the invention

[0005] In order to address the shortcomings of the prior art, the present invention proposes an active anti-money laundering early warning method and system based on a generative large language model, which utilizes the logical reasoning and generalization capabilities of the large model to realize functions such as suspicious account scanning, suspicious transaction early warning, and case-related transaction scanning. It discovers anomalies in massive transactions based on the natural language description of money laundering crimes. Compared with the traditional commercial bank anti-money laundering system, it realizes the full process automation and intelligence of the three stages of pre-prevention, in-process early warning, and post-inspection, greatly improving the real-time, accuracy and execution efficiency of the bank's anti-money laundering system.

[0006] To achieve the above objectives, the technical solutions adopted by the present invention include:

[0007] An active anti-money laundering early warning method based on a generative large language model, characterized by comprising:

[0008] S1. Establish a prompt word set, wherein the prompt word set includes a number of detection prompt words that are associated or independent of each other;

[0009] S2. Generate a simulated transaction data set based on the prompt word set using a preset large language model;

[0010] S3. Use the simulated transaction data set as a training set and a validation set to perform model training operations and generate a rule scanning model;

[0011] S4. Acquire real-time transaction data, and use a preset large language model to perform a suspicious prejudgment operation on the real-time transaction data to determine whether a transaction item corresponding to the real-time transaction data is a high-risk item;

[0012] S5. When determining whether the transaction item corresponding to the real-time transaction data is a high-risk item, calling the rule scanning model to perform a risk scanning operation on the transaction item corresponding to the real-time transaction data to determine whether the transaction item has risk factors;

[0013] S6. When it is determined that a transaction item has risk factors, risk warning information for the corresponding transaction item is generated.

[0014] Furthermore, the step S2 comprises:

[0015] Combining detection prompt words with associated relationships to generate a simulated transaction data set;

[0016] And / or, combining mutually independent detection prompt words to generate a simulated transaction data set.

[0017] Furthermore, the rule scanning model includes a suspicious account rule scanning model and a suspicious transaction rule scanning model.

[0018] Furthermore, the calling rule scanning model to perform a risk scanning operation on the transaction items corresponding to the real-time transaction data includes:

[0019] Determine whether the transaction account involved in the transaction item is an existing suspicious account;

[0020] When it is determined that the transaction account involved in the transaction item does not belong to the existing suspicious account, the suspicious transaction rule scanning model is called to perform a risk scanning operation;

[0021] When it is determined that the transaction account involved in the transaction item belongs to an existing suspicious account, the suspicious account rule scanning model and the suspicious transaction rule scanning model are called to perform risk scanning operations respectively.

[0022] Furthermore, the using of a preset large language model to perform a suspiciousness prediction operation on the real-time transaction data includes:

[0023] Convert real-time transaction data into information to be checked based on preset check templates;

[0024] The information to be checked is input into the large language model to perform suspicious prediction operations.

[0025] Furthermore, the method further comprises:

[0026] Matching the real-time transaction data and the information to be checked according to the prompt word set, and determining whether the information to be checked lacks the detection prompt word;

[0027] When it is determined that the information to be checked lacks the detection prompt word, the inspection template is replaced, and the real-time transaction data is converted into the information to be checked using the replaced inspection template.

[0028] The present invention also relates to an active anti-money laundering early warning system based on a generative large language model, which is characterized by comprising:

[0029] A prompt word management module, used to establish a prompt word set;

[0030] A simulation data management module, used to generate a simulation transaction data set according to a prompt word set using a preset large language model;

[0031] A rule scanning model training module is used to perform model training operations using a simulated transaction data set as a training set and a validation set to generate a rule scanning model;

[0032] A prediction module, used to use a preset large language model to perform a suspicious prediction operation on the real-time transaction data to determine whether the transaction item corresponding to the real-time transaction data belongs to a high-risk item;

[0033] A scanning execution module is used to call the rule scanning model to perform risk scanning operations on the transaction items corresponding to the real-time transaction data to determine whether there are risk factors in the transaction items;

[0034] The early warning module is used to generate risk early warning information for the corresponding transaction item when it is determined that there are risk factors in the transaction item.

[0035] The present invention also relates to a computer-readable storage medium, characterized in that a computer program is stored on the storage medium, and the computer program implements the above method when executed by a processor.

[0036] The present invention also relates to an electronic device, characterized in that it comprises a processor and a memory;

[0037] The memory is used to store the prompt word set, the simulated transaction data set and the real-time transaction data;

[0038] The processor is used to execute the above method by calling the prompt word set, the simulated transaction data set and the real-time transaction data.

[0039] The present invention also relates to a computer program product, comprising a computer program and / or instructions, characterized in that the computer program and / or instructions implement the steps of the above method when executed by a processor.

[0040] The beneficial effects of the present invention are:

[0041] The active anti-money laundering early warning method and system based on the generative large language model of the present invention are adopted, and the logical reasoning ability and generalization ability of the large model are utilized to realize the functions of suspicious account scanning, suspicious transaction early warning, and case-related transaction scanning. According to the natural language description of money laundering crimes, the accounts, transaction details, and the gangs involved in the case are inferred, and abnormal points in the massive transactions are found. Compared with the traditional commercial bank anti-money laundering system, the full process automation and intelligence of the three stages of pre-prevention, in-process early warning, and post-inspection are realized, which greatly improves the real-time, accuracy, and execution efficiency of the bank's anti-money laundering system. The active anti-money laundering early warning method and system of the present invention can automatically complete the screening of suspicious accounts, transactions, and blacklists, and finally give early warnings for suspicious transactions, and can continuously improve the judgment ability of the intelligent core for anti-money laundering cases in combination with the human feedback reinforcement learning (RLHF) method. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 The figure is a flow chart of the active anti-money laundering early warning method based on the generative large language model of the present invention.

[0043] Figure 2 The figure is a schematic diagram of the structure of the active anti-money laundering early warning system based on the generative large language model of the present invention. DETAILED DESCRIPTION

[0044] In order to more clearly understand the content of the present invention, it will be described in detail with reference to the accompanying drawings and embodiments. It should be noted that in the embodiments of the present application, some existing solutions in the industry such as certain software, components, models, etc. may be mentioned, which should be considered as exemplary, and their purpose is only to illustrate the feasibility of the implementation of the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0045] The acquisition, transmission, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0046] The first aspect of the present invention relates to a process of steps as follows Figure 1 The active anti-money laundering early warning method based on the generative large language model shown includes:

[0047] S1. Establish a prompt word set, wherein the prompt word set includes a plurality of detection prompt words that are associated or independent of each other.

[0048] In the specific implementation, the bank data warehouse is mainly used to extract and save the text data, historical transaction data, account-related feature data required for model training, etc., and is responsible for storing and managing the prompt words for the execution of large language model tasks, and preferably gives prompt word effect scores for the execution effect. For example, in the pre-prevention stage, according to the Chinese description prompt words of suspicious account characteristics entered by the administrator, such as transaction time in the early morning, multiple small-amount trials below 10 yuan, account balance is 0 for a long time but remains active and does not sleep, etc., simulated transaction data is generated; in the early warning stage, according to the Chinese description prompt words of the current suspicious transaction characteristics entered by the administrator, such as multiple transactions in a short period of time, whether it is on the blacklist, whether it is a transaction with a blacklist account, fast in and out in a short period of time, large-amount in and out that does not match the daily transaction level, etc.; in the post-inspection stage, according to the Chinese description prompt words of suspicious transaction characteristics entered by the administrator, such as multiple users logged in to the transaction device, abnormal transaction object account, strangers involved in the transaction counterparty and the existence of fund collection, etc., simulated transaction data are generated. By optimizing the prompt word engineering, the large language model is guided to understand the logical key between suspicious transaction characteristics, and then judge the suspiciousness of the transaction. For example, if the same mobile device ID logs into multiple mobile banking accounts and conducts multiple transfer transactions within a short period of time, the logical reasoning ability of the large language model can determine whether there is a suspicious logical connection between the above three features: a single device, logging into multiple accounts, and multiple transactions; and the generality of the large language model can automatically determine whether conditions such as a short period of time, multiple mobile banking accounts, and multiple transfer transactions have met the triggering conditions, rather than manually setting thresholds.

[0049] S2. Use the preset large language model to generate a simulated transaction data set based on the prompt word set.

[0050] Preferably, it includes combining detection prompt words with association relationship to generate a simulated transaction data set; and / or combining detection prompt words independent of each other to generate a simulated transaction data set.

[0051] In specific implementation, the large language model can be based on the large language model of the existing technology or the pre-trained model of the open source large language model, and can be adjusted according to the actual usage (for example, the adjustment method based on LORA). At the same time, the large language model can also be optimized by bringing it into the reinforcement learning module after marking the results of the manual verification of suspicious transactions.

[0052] S3. Use the simulated transaction data set as the training set and the validation set to perform model training operations and generate a rule scanning model.

[0053] Preferably, the rule scanning model includes a suspicious account rule scanning model and a suspicious transaction rule scanning model.

[0054] In specific implementation, the suspicious account rule scanning model is responsible for regularly scanning the bank's existing suspicious accounts, including small-amount trials, IP anomalies, account sharing, long-term lurking, and association with conversions of accounts involved in the case; the suspicious transaction rule scanning model is responsible for regularly scanning the suspiciousness of transactions that have occurred in the bank, including transaction purpose, capital flow, capital collection, quick in and out, etc.

[0055] S4. Acquire real-time transaction data, and use a preset large language model to perform a suspicious prediction operation on the real-time transaction data to determine whether a transaction item corresponding to the real-time transaction data is a high-risk item.

[0056] Preferably, executing the suspiciousness prediction operation includes converting the real-time transaction data into information to be checked according to a preset check template; and inputting the information to be checked into the large language model to execute the suspiciousness prediction operation.

[0057] Further preferably, it also includes matching the real-time transaction data and the information to be checked based on the prompt word set, and determining whether the information to be checked lacks the detection prompt word; when it is determined that the information to be checked lacks the detection prompt word, replacing the inspection template, and using the replaced inspection template to convert the real-time transaction data into the information to be checked.

[0058] In the specific implementation, the bank's real-time transaction data platform can be selected to store the bank's real-time transaction data and extract key features. Based on the prompt words describing the suspicious transaction features, the large language model can understand the logical relationship between the suspicious transaction features and the feature matrix of the suspicious transaction; the transaction details to be checked (information to be checked) are input into the large language model one by one, and the large language model determines the execution suspicion one by one. If the result is safe, it will be skipped; if the result is suspicious, the current transaction will be blocked, and the rule scanning model will review the model's suspicion.

[0059] S5. When determining whether a transaction item corresponding to the real-time transaction data is a high-risk item, the rule scanning model is called to perform a risk scanning operation on the transaction item corresponding to the real-time transaction data to determine whether the transaction item has risk factors.

[0060] Preferably, it includes judging whether the transaction account involved in the transaction item belongs to an existing suspicious account; when it is judged that the transaction account involved in the transaction item does not belong to an existing suspicious account, calling the suspicious transaction rule scanning model to perform a risk scanning operation; when it is judged that the transaction account involved in the transaction item belongs to an existing suspicious account, calling the suspicious account rule scanning model and the suspicious transaction rule scanning model to perform risk scanning operations respectively.

[0061] S6. When it is determined that a transaction item has risk factors, risk warning information for the corresponding transaction item is generated.

[0062] Preferably, an operator may choose to manually review and pre-judge the transaction with the transaction initiator within a preset time (eg, within 2 hours).

[0063] When the above method is actually applied, it is preferred to establish rule scanning models according to different stages of anti-money laundering early warning work (detection prompt words corresponding to different stages) to perform inspection and scanning operations. For example, in the pre-prevention stage, the corresponding rule scanning model is obtained by training after the simulated transaction data is generated, and the model is executed regularly (can be set to weekly) to scan bank accounts and generate a list of suspicious accounts and feed it back to the administrator for manual verification; in the in-process early warning stage, the saved suspicious transaction screening prompt words are combined with the implementation transaction feature data to generate multiple rounds of questions and answers for real-time transactions, and a suspicious judgment on the current transaction is obtained, and the judgment result is saved and submitted to subsequent manual verification; in the post-inspection stage, the corresponding rule scanning model is obtained by training after the simulated transaction data is generated, and the model is executed regularly (for example, set to weekly) to scan the bank's existing transactions and generate a list of suspicious accounts and feed it back to the administrator for manual verification.

[0064] Another aspect of the present invention also relates to an active anti-money laundering early warning system based on a generative large language model, the structure of which is as follows: Figure 2 As shown, including:

[0065] A prompt word management module, used to establish a prompt word set;

[0066] A simulation data management module, used to generate a simulation transaction data set according to a prompt word set using a preset large language model;

[0067] A rule scanning model training module is used to perform model training operations using a simulated transaction data set as a training set and a validation set to generate a rule scanning model;

[0068] A prediction module, used to use a preset large language model to perform a suspicious prediction operation on the real-time transaction data to determine whether the transaction item corresponding to the real-time transaction data belongs to a high-risk item;

[0069] A scanning execution module is used to call the rule scanning model to perform risk scanning operations on the transaction items corresponding to the real-time transaction data to determine whether there are risk factors in the transaction items;

[0070] The early warning module is used to generate risk early warning information for the corresponding transaction item when it is determined that there are risk factors in the transaction item.

[0071] By using this system, the above-mentioned operation processing method can be executed and the corresponding technical effects can be achieved.

[0072] An embodiment of the present invention also provides a computer-readable storage medium capable of implementing all the steps in the method in the above embodiment, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, all the steps in the method in the above embodiment are implemented.

[0073] An embodiment of the present invention also provides an electronic device for executing the above method. As an implementation device of the method, the electronic device at least has a processor and a memory, and in particular, the memory stores data and related computer programs required for executing the method, such as a prompt word set, a simulated transaction data set and real-time transaction data, etc., and the processor calls the data and program in the memory to execute all steps of the implementation method and obtains the corresponding technical effect.

[0074] Preferably, the electronic device may include a bus architecture, the bus may include any number of interconnected buses and bridges, and the bus will include various circuits linked together by one or more processors and memories. The bus may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are all well known in the art and are therefore not further described herein. The bus interface provides an interface between the bus and the receiver and transmitter. The receiver and transmitter may be the same component, i.e., a transceiver, providing a unit for communicating with various other systems over a transmission medium. The processor is responsible for managing the bus and general processing, while the memory may be used to store data used by the processor when performing operations.

[0075] In addition, the electronic device may further include components such as a communication module, an input unit, an audio processor, a display, and a power supply. The processor (or controller, operation control) used may include a microprocessor or other processor device and / or a logic device, which receives input and controls the operation of various components of the electronic device; the memory may be one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory, or other suitable devices, which may store the above-mentioned related data information, and may also store a program for executing related information, and the processor may execute the program stored in the memory to realize information storage or processing, etc.; the input unit is used to provide input to the processor, for example, it may be a key or a touch input device; the power supply is used to provide power to the electronic device; the display is used to display display objects such as images and text, for example, it may be an LCD display. The communication module is a transmitter / receiver that sends and receives signals via an antenna. The communication module (transmitter / receiver) is coupled to the processor to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal. Based on different communication technologies, multiple communication modules may be provided in the same electronic device, such as a cellular network module, a Bluetooth module and / or a wireless local area network module, etc. The communication module (transmitter / receiver) is also coupled to a speaker and a microphone via an audio processor to provide an audio output via the speaker and receive an audio input from the microphone, thereby realizing a common telecommunication function. The audio processor may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor is also coupled to a central processor, so that recording can be performed on the machine through the microphone, and the sound stored on the machine can be played through the speaker.

[0076] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0077] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or multiple boxes.

[0078] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction system, which is implemented in the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0079] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps of the functions specified in a box or multiple boxes. Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they know the basic creative concept. Therefore, the attached claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0080] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. An active anti-money laundering early warning method based on a generative large language model, characterized in that: include: S1. Establish a prompt word set, wherein the prompt word set includes a number of detection prompt words that are associated or independent of each other; S2. Generate a simulated transaction data set based on the prompt word set using a preset large language model; S3. Use the simulated transaction data set as a training set and a validation set to perform model training operations and generate a rule scanning model; S4. Acquire real-time transaction data, and use a preset large language model to perform a suspicious prejudgment operation on the real-time transaction data to determine whether a transaction item corresponding to the real-time transaction data is a high-risk item; S5. When determining whether the transaction item corresponding to the real-time transaction data is a high-risk item, calling the rule scanning model to perform a risk scanning operation on the transaction item corresponding to the real-time transaction data to determine whether the transaction item has risk factors; S6. When it is determined that a transaction item has risk factors, risk warning information for the corresponding transaction item is generated.

2. The method according to claim 1, characterized in that The step S2 comprises: Combining detection prompt words with associated relationships to generate a simulated transaction data set; And / or, combining mutually independent detection prompt words to generate a simulated transaction data set.

3. The method according to claim 1, characterized in that The rule scanning model includes a suspicious account rule scanning model and a suspicious transaction rule scanning model.

4. The method according to claim 3, characterized in that The calling rule scanning model to perform risk scanning operations on transaction items corresponding to real-time transaction data includes: Determine whether the transaction account involved in the transaction item is an existing suspicious account; When it is determined that the transaction account involved in the transaction item does not belong to the existing suspicious account, the suspicious transaction rule scanning model is called to perform a risk scanning operation; When it is determined that the transaction account involved in the transaction item belongs to an existing suspicious account, the suspicious account rule scanning model and the suspicious transaction rule scanning model are called to perform risk scanning operations respectively.

5. The method according to claim 1, characterized in that The using of a preset large language model to perform a suspicious prediction operation on the real-time transaction data includes: Convert real-time transaction data into information to be checked based on preset check templates; The information to be checked is input into the large language model to perform suspicious prediction operations.

6. The method according to claim 5, characterized in that The method further comprises: Matching the real-time transaction data and the information to be checked according to the prompt word set, and determining whether the information to be checked lacks the detection prompt word; When it is determined that the information to be checked lacks the detection prompt word, the inspection template is replaced, and the real-time transaction data is converted into the information to be checked using the replaced inspection template.

7. An active anti-money laundering early warning system based on a generative large language model, characterized in that: include: A prompt word management module, used to establish a prompt word set; A simulation data management module, used to generate a simulation transaction data set according to a prompt word set using a preset large language model; A rule scanning model training module is used to perform model training operations using a simulated transaction data set as a training set and a validation set to generate a rule scanning model; A prediction module, used to use a preset large language model to perform a suspicious prediction operation on the real-time transaction data to determine whether the transaction item corresponding to the real-time transaction data belongs to a high-risk item; A scanning execution module is used to call the rule scanning model to perform risk scanning operations on the transaction items corresponding to the real-time transaction data to determine whether there are risk factors in the transaction items; The early warning module is used to generate risk early warning information for the corresponding transaction item when it is determined that there are risk factors in the transaction item.

8. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. An electronic device, characterized in that: including a processor and a memory; The memory is used to store the prompt word set, the simulated transaction data set and the real-time transaction data; The processor is used to execute the method according to any one of claims 1 to 6 by calling a prompt word set, a simulated transaction data set and real-time transaction data.

10. A computer program product comprising a computer program and / or instructions, characterized in that: When the computer program and / or the instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Citation Information

Cited By

  • Risk control behavior early warning prediction method and system based on Al identification

    CN119379427A

  • Anti-money laundering risk rule generation method and system based on multi-modal large language model

    CN120598567A

  • Financial transaction risk analysis method and system based on multi-technology cooperation, medium, terminal and program product

    CN122023005A