CN-122023005-A - Financial transaction risk analysis method, system, medium, terminal and program product based on multi-technology collaboration

CN122023005ACN 122023005 ACN122023005 ACN 122023005ACN-122023005-A

Abstract

The application provides a financial transaction risk analysis method, a system, a medium, a terminal and a program product based on multi-technology collaboration, wherein the method comprises the steps of obtaining internal structured data and external structured data related to a target customer financial transaction; the method comprises the steps of extracting risk factors in each risk domain from internal structured data and external structured data to form risk features, inputting the risk features into a risk dimension identification model to output each risk domain score, calculating contribution scores of each risk factor to the risk domain scores based on SHAP analysis technology to generate a plurality of risk factors with the highest contribution scores, inputting each risk domain score, the plurality of risk factors with the highest contribution scores and corresponding original data into a large language model, and generating a financial transaction risk analysis report by the large language model according to a prompt word template. The application can accurately focus the risk focus of the financial transaction of the target customer and produce a high-quality and accurate financial transaction risk analysis report.

Inventors

Kuang Lanjuan
ZHU XIAOYUN
LI BAOJUN
LIN CHEN
HUANG SHANHE
HUANG YUZHU
ZHANG JIE
CHEN JIAHUI
SHEN ZHONGWU
YANG MINGMING
ZHU SHENGHUI
YU MENGFEI

Assignees

友邦人寿保险有限公司

Dates

Publication Date: 20260512
Application Date: 20260214

Claims (10)

1. A financial transaction risk analysis method based on multi-technology collaboration, comprising: acquiring internal structured data related to a target customer financial transaction, acquiring related external unstructured data from an external source based on a robot process automation technology, and preprocessing the acquired external unstructured data to obtain external structured data; pre-constructing a risk domain formed by a plurality of risk factors according to a back money laundering supervision rule, and extracting the risk factors in each risk domain from the internal structural data and the external structural data to form a risk feature; Inputting the risk characteristics into a pre-constructed risk dimension identification model, and outputting each risk domain score; calculating contribution scores of all risk factors to the risk domain scores based on SHAP analysis technology, and generating a plurality of risk factors with the highest contribution scores; And inputting the risk factors with the highest risk domain scores and contribution scores, the corresponding internal structured data and the corresponding external structured data into a large language model, and generating a financial transaction risk analysis report by the large language model according to a pre-constructed prompt word template.
2. The financial transaction risk analysis method based on multi-technology collaboration according to claim 1, wherein the pre-construction process of the risk dimension identification model is as follows: collecting cases marked as suspicious transactions historically, defining the cases reported to a money laundering supervision organization after manual research and judgment as positive samples, and defining the cases not reported after manual rechecking as negative samples, so as to construct a supervision learning label set; Performing standardization processing on risk features extracted from the internal structured data and the external unstructured data to obtain a risk feature matrix; inputting the constructed supervised learning label set and the risk feature matrix into a logistic regression model for training; constructing a negative log-likelihood loss function based on maximum likelihood estimation and limiting the complexity of the logistic regression model using L1 or L2 regularization; and minimizing the negative log likelihood loss function by using a gradient descent method to obtain optimal parameters so as to construct and obtain the risk dimension identification model.
3. The financial transaction risk analysis method based on multi-technology collaboration according to claim 1, wherein the pre-construction process of the prompt word template is as follows: acquiring a plurality of expert written reports, and desensitizing the acquired expert written reports; inputting the expert writing report after desensitization treatment into a large language model, and refining an analysis template corresponding to each risk domain; And converting the extracted analysis templates into prompt word templates corresponding to the risk domains and storing the prompt word templates to form a prompt word template library.
4. The method for analyzing risk of financial transaction based on multi-technology collaboration according to claim 1, wherein the preprocessing of the obtained external unstructured data to obtain external structured data comprises the following specific steps: calling an OCR engine to extract text content and corresponding position information from webpage screenshot data and PDF files which are acquired from the outside; Inputting the webpage screenshot data, the PDF file, the text content extracted by the OCR engine and the corresponding position information into a multi-modal large language model, and outputting external structured data by the multi-modal large language model according to a preset prompt word.
5. The method of claim 1, further comprising inputting feedback information for auditing, modifying, and validating the generated financial transaction risk analysis report based on the compliance person into a large language model for iterative optimization.
6. The multi-technology collaboration-based financial transaction risk analysis method of claim 1, wherein the risk domain includes customer identity risk information, policy information, transaction behavior, family risk information, fund anomaly information, and external risk information.
7. A financial transaction risk analysis system based on multi-technology collaboration, comprising: The acquisition module is used for acquiring internal structured data related to the financial transaction of the target customer, acquiring related external unstructured data from an external source based on a robot process automation technology, and preprocessing the acquired external unstructured data to obtain the external structured data; the feature extraction module is used for pre-constructing a risk domain formed by a plurality of risk factors according to a money back-washing supervision rule, and extracting the risk factors in each risk domain from the internal structural data and the external structural data to form risk features; the scoring module is used for inputting the risk characteristics into a pre-constructed risk dimension identification model and outputting each risk domain score, calculating contribution scores of each risk factor to the risk domain score based on SHAP analysis technology, and generating a plurality of risk factors with the highest contribution scores; The report generation module is used for inputting each risk domain score, a plurality of risk factors with highest contribution score, corresponding internal structured data and corresponding external structured data into a large language model, and the large language model generates a financial transaction risk analysis report according to a pre-constructed prompt word template.
8. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the method according to any one of claims 1 to 6.
9. A computer program product comprising computer program code means for causing a computer to carry out the method as claimed in any one of claims 1 to 6 when said computer program code means are run on the computer.
10. An electronic terminal comprising a memory, a processor and a computer program stored on the memory, characterized in that the processor executes the computer program to implement the method according to any of claims 1 to 6.

Description

Financial transaction risk analysis method, system, medium, terminal and program product based on multi-technology collaboration Technical Field The application relates to the technical field of artificial intelligence, in particular to a financial transaction risk analysis method, a financial transaction risk analysis system, a financial transaction risk analysis medium, a financial transaction risk analysis terminal and a financial transaction risk analysis program product based on multi-technology collaboration. Background Under the background of increasingly strict financial compliance supervision, financial institutions need to conduct risk screening on massive customer transaction behaviors. The conventional method relies on a rule engine to grasp suspicious transactions and then manually combines internal and external information to write risk reports, but the process has the problems that the judgment of transaction risks needs to combine various data sources, including internal transaction data, external data such as executed information of clients and the like, the acquisition of the external information needs manual inquiry by one person, time and labor are consumed, risk judgment is highly dependent on expert experience, a new person is difficult to quickly master complex analysis logic and supervision points, and in addition, the time and labor are consumed to write risk assessment reports, and each report needs to take about 6 hours on average, so that a large amount of human resources are wasted, and the high concurrency screening requirement is difficult to meet. To address the above challenges, with the development and wide application of large language models (Large Language Model, LLM) of generated AI, the industry began to explore the introduction of generated AI technology to build an intelligent risk identification and report generation system. However, while the generated AI scheme has significant potential in terms of risk analysis logic and report writing, the following key technical difficulties remain faced in the actual landing process: (1) External data such as an execution network, hundred degrees and other websites are difficult to quickly acquire, and have no data interface for docking, so that the data acquisition difficulty is high. (2) If raw data such as unprocessed transaction records, policy lists and the like are directly input into the LLM, the model is easy to generate fuzzy, generalized and even wrong output due to the lack of clear risk focuses and business contexts, and the severe requirements of financial scenes on accuracy and interpretability are difficult to meet. (3) The expert writes report experience which is difficult to copy, the high-quality risk report depends on comprehensive judgment of compliance expert on supervision rules, transaction abnormality and business background, the experience is mostly implicit knowledge, structured precipitation is difficult, and even if a new person grasps a process, key points or expression is often omitted due to lack of judgment, so that report quality is unstable, culture cost is high, and large-scale difficulty is caused. Accordingly, there is a need for a financial transaction risk analysis method, system, medium, terminal and program product based on multi-technology collaboration, which solve the above-mentioned problems in the prior art. Disclosure of Invention In view of the above-mentioned drawbacks of the prior art, the present application is directed to a financial transaction risk analysis method, system, medium, terminal and program product based on multi-technology collaboration, which are used for solving the technical problems that in the prior art, external data are difficult to obtain quickly, uncertainty exists in LLM direct analysis of original structured data, and expertise of writing reports by experts is difficult to copy. To achieve the above and other related objects, a first aspect of the present application provides a financial transaction risk analysis method based on multi-technology collaboration, which includes obtaining internal structured data related to a target customer financial transaction, obtaining external unstructured data related to the target customer financial transaction from an external source based on a robotic process automation technology, preprocessing the obtained external unstructured data to obtain external structured data, pre-constructing a risk domain composed of a plurality of risk factors according to a back-washing supervision rule, extracting risk factors in each risk domain from the internal structured data and the external structured data to form risk features, inputting the risk features into a pre-constructed risk dimension recognition model, outputting each risk domain score, calculating contribution points of each risk factor to the risk domain score based on an SHAP analysis technology, generating a plurality of risk factors with highest contribution p