Financial transaction fraud risk prediction method and system

Abstract

The invention discloses a financial transaction fraud risk prediction method and system, and the method comprises the steps: collecting and preprocessing financial transaction data, and obtaining the processed data; constructing a risk prediction model based on the processed data; optimizing the risk prediction model to obtain a final model; and completing risk prediction based on the final model. According to the method, the fraud risk knowledge graph is constructed, and the random walk algorithm and the clustering algorithm are utilized to carry out risk scoring and group division on the transaction entities, so that fraud behaviors in financial transactions can be effectively identified and predicted, and the accuracy and coverage rate of fraud identification are improved. Meanwhile, the method and the system can adapt to dynamic changes of transaction modes and fraud skills, and through regularly extracting new fraud transaction samples from real-time data streams, risk rules and prediction models are continuously optimized and updated, so that efficient monitoring and early warning of financial transaction fraud behaviors are realized.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method and system for predicting financial transaction fraud risks. Background Art

[0002] With the rapid expansion of the global financial market and the deepening of digital transformation, financial transaction activities are becoming more frequent and transaction methods are becoming more diverse. However, financial transaction fraud has also increased, causing huge economic losses to financial institutions and consumers. Traditional fraud detection methods mainly rely on rule engines and expert experience, which are unable to cope with complex and changeable fraud patterns. Especially in the context of the big data era, the amount of financial transaction data has surged, the data sources are diverse, and the fraud methods are constantly updated, which greatly increases the difficulty of fraud detection. Existing fraud detection systems often lack the ability to mine deep relationships in transaction data and cannot effectively identify complex fraud behaviors across multiple entities and channels. Therefore, developing a financial transaction fraud risk prediction method and system that can process large-scale data, identify complex relationship networks, and update prediction models in real time is of great significance to improving the security of financial transactions and reducing fraud losses. This technology needs to be able to adapt to the dynamic changes in transaction patterns and fraud methods, and improve the accuracy and coverage of fraud identification through continuous learning and optimization to meet the increasingly severe financial security challenges. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a method for predicting financial transaction fraud risk, the method comprising:

[0004] Collect financial transaction data and pre-process it to obtain processed data;

[0005] Based on the processed data, construct a risk prediction model;

[0006] Optimizing the risk prediction model to obtain a final model;

[0007] Based on the final model, risk prediction is completed.

[0008] Preferably, an ETL tool is used to clean, transform and load the acquired financial transaction data to obtain the processed data; the method includes: performing quality inspection on the acquired financial transaction data according to pre-established data quality rules to form a data quality report; and using a data cleaning method to process the quality problem data identified in the data quality report.

[0009] Preferably, based on the processed data, a fraud risk knowledge graph is constructed using graph database technology, and a global fraud risk view is formed by mining the complex relationship network between transaction entities; when the identified association relationship meets the preset fraud risk judgment rules, the transaction entity is judged to have a fraud risk.

[0010] Preferably, according to the processed data, based on the constructed risk knowledge graph, a random walk algorithm is used to calculate the fraud risk score of each transaction entity; and transaction entities with similar fraud risk scores are clustered through a clustering algorithm to obtain a group of transaction entities with similar fraud risks.

[0011] Preferably, based on the processed data, in response to the dynamic changes in transaction patterns and fraud methods, new fraud transaction samples are periodically extracted from real-time data streams, and the existing risk rules and risk prediction models are continuously optimized and updated to improve the accuracy and coverage of fraud identification, thereby obtaining the final model.

[0012] The present invention also provides a financial transaction fraud risk prediction system, the system is used to implement the above method, including: a collection module, a construction module, an optimization module and a prediction module;

[0013] The acquisition module is used to collect financial transaction data and perform preprocessing to obtain processed data;

[0014] The construction module is used to construct a risk prediction model based on the processed data;

[0015] The optimization module is used to optimize the risk prediction model to obtain a final model;

[0016] The prediction module is used to complete risk prediction based on the final model.

[0017] Preferably, the acquisition module uses ETL tools to clean, convert and load the acquired financial transaction data to obtain the processed data; the method includes: performing quality inspection on the acquired financial transaction data according to pre-established data quality rules to form a data quality report; and using a data cleaning method to process the quality problem data identified in the data quality report.

[0018] Preferably, the construction module constructs a fraud risk knowledge graph based on the processed data using graph database technology, and forms a global fraud risk view by mining the complex relationship network between transaction entities; when the identified association relationship meets the preset fraud risk judgment rules, the transaction entity is judged to have a fraud risk.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] The present invention can effectively identify and predict fraudulent behavior in financial transactions and improve the accuracy and coverage of fraud identification by constructing a fraud risk knowledge graph and using a random walk algorithm and a clustering algorithm to perform risk scoring and group division on transaction entities. At the same time, the method and system can adapt to the dynamic changes in transaction patterns and fraudulent methods, and continuously optimize and update risk rules and prediction models by regularly extracting new fraudulent transaction samples from real-time data streams, thereby achieving efficient monitoring and early warning of financial transaction fraud, reducing economic losses of financial institutions and consumers, and enhancing the security and reliability of financial transactions. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0022] Figure 1 The figure is a schematic diagram of a method flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present disclosure should be understood by people with ordinary skills in the field to which the present disclosure belongs. The "first", "second" and similar words used in the embodiments of the present disclosure do not indicate any order, quantity or importance, but are only used to distinguish different components. "Including" or "comprising" and similar words mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0025] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] Embodiment 1

[0027] like Figure 1 FIG. 1 is a schematic diagram of the method flow of this embodiment, and the steps include:

[0028] S1. Collect financial transaction data and pre-process it to obtain processed data.

[0029] ETL tools are used to clean, convert and load financial transaction data to obtain processed data, providing a high-quality data foundation for subsequent fraud risk analysis. According to pre-established data quality rules, the quality of the acquired financial transaction data is tested to identify noise data, duplicate data, inconsistent data, etc., and form a data quality report. For the quality problem data identified in the data quality report, the corresponding data cleaning method is used to process it. In addition, character data needs to be converted to numeric data, date data needs to be converted to timestamp format, and fields with the same name from different sources need to be mapped to a unified field name.

[0030] Specifically, data quality rules are the basis of data governance. This embodiment defines rules: the user's age must be greater than 0 and less than 150, the transaction amount must be greater than 0, and the phone number must conform to a specific format. These rules are used to detect problems in the acquired financial transaction data. Assume that user information and transaction data are obtained from an e-commerce platform, an offline store, and a third-party payment platform. After data quality detection, some problems were found: the age of some users is missing or entered incorrectly (such as a negative age), the amount of some transaction records is 0, the phone number format of some users is incorrect, and some users' information is repeated or inconsistent in different data sources (for example, the same user's name has different spellings in different data sources). These problematic data will form a data quality report. For example, the report points out that the missing rate of the age field of the e-commerce platform data source is 5%, and there are 100 duplicate transaction records in the offline store data source. Different data cleaning algorithms are used for data quality reports. For example, if the user's age is missing, it can be supplemented with the nearest neighbor interpolation algorithm. Assuming that a user's age is missing, but his consumption habits are very similar to those of another user who is 25 years old, the user's age can be supplemented to 25 years old. For noisy data, such as a user's transaction amount being recorded as an abnormally large value due to a system failure, the Kalman filter algorithm can be used to remove noise and make it close to the true value. For inconsistent data, such as the same user's name in different data sources is different, data association analysis can be performed, such as comparing the user's ID number, mobile phone number and other information to determine the unique and accurate name. After data cleaning, format conversion and field mapping are required. For example, convert the user's gender field from "male" and "female" to 1 and 0, and convert date data to timestamp format, such as "2024-07-27" to 1701024000. Fields with the same name from different sources, such as "user name" and "transaction time", are mapped to unified field names, such as "user_name" and "transaction_time".

[0031] S2. Build a risk prediction model based on the processed data.

[0032] Based on the processed data, the fraud risk knowledge graph is constructed using graph database technology. By mining the complex relationship network between transaction entities, a global fraud risk view is formed. If the identified association relationship meets the preset fraud risk judgment rules, the transaction entity is judged to have a fraud risk. Specifically, the core of constructing the fraud risk knowledge graph is to use graph database technology to display and analyze the complex relationship between transaction entities. For example, users, mobile phone numbers, IP addresses, devices, transaction orders, etc. can be regarded as nodes in the graph, and the relationship between them, such as users using mobile phone numbers, users using devices to place orders, and transaction orders using IP addresses, etc. can be regarded as edges in the graph. In this way, the association between the various entities can be clearly seen. If a user is associated with multiple high-risk mobile phone numbers or IP addresses, the risk score of the user itself will also increase accordingly. Prediction models use association rule mining to help discover patterns of fraudulent behavior; for example, it may be found that some fraudulent transactions often occur late at night, use virtual IP addresses, and have large transaction amounts. These rules can be used to identify new suspected fraudulent transactions.

[0033] Then, based on the entity relationships in the fraud risk knowledge graph, the random walk algorithm is used to calculate the fraud risk score of each transaction entity. The transaction entities with similar fraud risk scores are clustered using a clustering algorithm to obtain a group of transaction entities with similar fraud risks.

[0034] Specifically, the model uses a random walk algorithm to calculate the fraud risk score of each transaction entity. When a risk propagates in the graph, nodes with high risk will propagate the risk to their adjacent nodes. Through iterative calculations, each node will eventually obtain a risk score that reflects its risk level in the entire network. For example, a device associated with multiple known fraud users will have a higher risk score. Clustering algorithms can group transaction entities with similar risks. For example, it can be found that a group of users use the same device, the same IP address segment, and have similar transaction behavior patterns, then this group of users may belong to the same fraud gang.

[0035] The model of this embodiment uses the unsupervised learning algorithm k-means to cluster transaction entities and classify transaction entities with similar fraud risk scores into the same cluster. For each cluster, the transaction behavior characteristics such as transaction frequency and transaction amount of the transaction entities in the cluster are analyzed to determine the fraud risk level of the cluster. For clusters with higher fraud risk levels, the transaction entities in the cluster are obtained, marked as high-risk transaction entities, and included in the key monitoring list.

[0036] Specifically, assume that the transaction entities are divided into three clusters: low-risk cluster, medium-risk cluster and high-risk cluster. For the low-risk cluster, the transaction frequency and transaction amount of the transaction entities are relatively low, for example, the number of transactions per month is less than 10 times, and the total transaction amount is less than 1,000 yuan. The transaction behavior characteristics of the transaction entities in the medium-risk cluster are between low risk and high risk, for example, the number of transactions per month is between 10 and 50 times, and the total transaction amount is between 1,000 yuan and 10,000 yuan. The transaction frequency and transaction amount of the transaction entities in the high-risk cluster are relatively high, for example, the number of transactions per month is greater than 50 times, and the total transaction amount is greater than 10,000 yuan. For each cluster, it is necessary to analyze the transaction behavior characteristics of the transaction entities in the cluster to determine the fraud risk level of the cluster. The k-means clustering algorithm includes:

[0037] 1. Randomly select k (3 in this embodiment) cluster centers to initialize each cluster.

[0038] 2. Traverse all data points and assign each data point to the cluster represented by the cluster center closest to it.

[0039] 3. Recalculate the cluster center of each cluster.

[0040] 4. Repeat steps 2 to 3 until convergence is reached.

[0041] S3. Optimize the risk prediction model to obtain the final model.

[0042] In response to the dynamic changes in transaction patterns and fraud methods, new fraud transaction samples are regularly extracted from real-time data streams, and existing risk rules and models are continuously optimized and updated to improve the accuracy and coverage of fraud identification.

[0043] According to the dynamic changes of transaction patterns and fraud methods, real-time transaction data streams are obtained, and new fraud transaction samples are extracted from them; the extracted new fraud transaction samples are analyzed to determine whether they comply with the existing risk rules. If not, new risk rules are generated based on the sample characteristics and added to the risk rule library; by adding new fraud transaction samples to the training set, the existing fraud identification model is incrementally trained and optimized so that the model can adapt to new fraud patterns.

[0044] For example, a new fraudulent transaction sample shows that the fraudster used the same IP address to conduct multiple large transactions in a short period of time, and the recipients of these transactions were all newly registered accounts. This may indicate that this is a new type of financial fraud. Generate new risk rules based on sample characteristics. For example, based on the above analysis, a new risk rule can be generated: "If the same IP address initiates more than 5 transactions exceeding 10,000 yuan within 1 hour, and the recipient account registration time is less than 7 days, then the transaction is high risk." This rule is added to the risk rule library for real-time monitoring.

[0045] S4. Complete risk prediction based on the final model.

[0046] It should be noted that the method of the embodiment of the present disclosure can be performed by a single device, such as a computer or a server. The method of the present embodiment can also be applied in a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present disclosure, and the multiple devices will interact with each other to complete the described method.

[0047] It should be noted that some embodiments of the present disclosure are described above. Other embodiments are within the scope of the attached claims. In some cases, it should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention. The actions or steps recorded in the claims can be performed in an order different from that in the above embodiment and still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0048] Embodiment 2

[0049] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present disclosure also provides a financial transaction fraud risk prediction system, including: a collection module, a construction module, an optimization module and a prediction module; the collection module is used to collect financial transaction data and perform preprocessing to obtain processed data; the construction module is used to construct a risk prediction model based on the processed data; the optimization module is used to optimize the risk prediction model to obtain a final model; the prediction module is used to complete risk prediction based on the final model. ,

[0050] The following will explain in detail how the present invention solves practical life problems in work in conjunction with this embodiment.

[0051] First, the collection module is used to collect financial transaction data and pre-process it to obtain the processed data.

[0052] ETL tools are used to clean, convert and load financial transaction data to obtain processed data, providing a high-quality data foundation for subsequent fraud risk analysis. According to pre-established data quality rules, the quality of the acquired financial transaction data is tested to identify noise data, duplicate data, inconsistent data, etc., and form a data quality report. For the quality problem data identified in the data quality report, the corresponding data cleaning method is used to process it. In addition, character data needs to be converted to numeric data, date data needs to be converted to timestamp format, and fields with the same name from different sources need to be mapped to a unified field name.

[0053] Specifically, data quality rules are the basis of data governance. This embodiment defines rules: the user's age must be greater than 0 and less than 150, the transaction amount must be greater than 0, and the phone number must conform to a specific format. These rules are used to detect problems in the acquired financial transaction data. Assume that user information and transaction data are obtained from an e-commerce platform, an offline store, and a third-party payment platform. After data quality detection, some problems were found: the age of some users is missing or entered incorrectly (such as a negative age), the amount of some transaction records is 0, the phone number format of some users is incorrect, and some users' information is repeated or inconsistent in different data sources (for example, the same user's name has different spellings in different data sources). These problematic data will form a data quality report. For example, the report points out that the missing rate of the age field of the e-commerce platform data source is 5%, and there are 100 duplicate transaction records in the offline store data source. Different data cleaning algorithms are used for data quality reports. For example, if the user's age is missing, it can be supplemented with the nearest neighbor interpolation algorithm. Assuming that a user's age is missing, but his consumption habits are very similar to those of another user who is 25 years old, the user's age can be supplemented to 25 years old. For noisy data, such as a user's transaction amount being recorded as an abnormally large value due to a system failure, the Kalman filter algorithm can be used to remove noise and make it close to the true value. For inconsistent data, such as the same user's name in different data sources is different, data association analysis can be performed, such as comparing the user's ID number, mobile phone number and other information to determine the unique and accurate name. After data cleaning, format conversion and field mapping are required. For example, convert the user's gender field from "male" and "female" to 1 and 0, and convert date data to timestamp format, such as "2024-07-27" to 1701024000. Fields with the same name from different sources, such as "user name" and "transaction time", are mapped to unified field names, such as "user_name" and "transaction_time".

[0054] The module then builds a risk prediction model based on the processed data.

[0055] Based on the processed data, the fraud risk knowledge graph is constructed using graph database technology. By mining the complex relationship network between transaction entities, a global fraud risk view is formed. If the identified association relationship meets the preset fraud risk judgment rules, the transaction entity is judged to have a fraud risk. Specifically, the core of constructing the fraud risk knowledge graph is to use graph database technology to display and analyze the complex relationship between transaction entities. For example, users, mobile phone numbers, IP addresses, devices, transaction orders, etc. can be regarded as nodes in the graph, and the relationship between them, such as users using mobile phone numbers, users using devices to place orders, and transaction orders using IP addresses, etc. can be regarded as edges in the graph. In this way, the association between the various entities can be clearly seen. If a user is associated with multiple high-risk mobile phone numbers or IP addresses, the risk score of the user itself will also increase accordingly. Prediction models use association rule mining to help discover patterns of fraudulent behavior; for example, it may be found that some fraudulent transactions often occur late at night, use virtual IP addresses, and have large transaction amounts. These rules can be used to identify new suspected fraudulent transactions.

[0056] Then, based on the entity relationships in the fraud risk knowledge graph, the random walk algorithm is used to calculate the fraud risk score of each transaction entity. The transaction entities with similar fraud risk scores are clustered using a clustering algorithm to obtain a group of transaction entities with similar fraud risks.

[0057] Specifically, the model uses a random walk algorithm to calculate the fraud risk score of each transaction entity. When a risk propagates in the graph, nodes with high risk will propagate the risk to their adjacent nodes. Through iterative calculations, each node will eventually obtain a risk score that reflects its risk level in the entire network. For example, a device associated with multiple known fraud users will have a higher risk score. Clustering algorithms can group transaction entities with similar risks. For example, it can be found that a group of users use the same device, the same IP address segment, and have similar transaction behavior patterns, then this group of users may belong to the same fraud gang.

[0058] The model of this embodiment uses the unsupervised learning algorithm k-means to cluster transaction entities and classify transaction entities with similar fraud risk scores into the same cluster. For each cluster, the transaction behavior characteristics such as transaction frequency and transaction amount of the transaction entities in the cluster are analyzed to determine the fraud risk level of the cluster. For clusters with higher fraud risk levels, the transaction entities in the cluster are obtained, marked as high-risk transaction entities, and included in the key monitoring list.

[0059] Specifically, assume that the transaction entities are divided into three clusters: low-risk cluster, medium-risk cluster and high-risk cluster. For the low-risk cluster, the transaction frequency and transaction amount of the transaction entities are relatively low, for example, the number of transactions per month is less than 10 times, and the total transaction amount is less than 1,000 yuan. The transaction behavior characteristics of the transaction entities in the medium-risk cluster are between low risk and high risk, for example, the number of transactions per month is between 10 and 50 times, and the total transaction amount is between 1,000 yuan and 10,000 yuan. The transaction frequency and transaction amount of the transaction entities in the high-risk cluster are relatively high, for example, the number of transactions per month is greater than 50 times, and the total transaction amount is greater than 10,000 yuan. For each cluster, it is necessary to analyze the transaction behavior characteristics of the transaction entities in the cluster to determine the fraud risk level of the cluster. The k-means clustering algorithm includes:

[0060] 1. Randomly select k (3 in this embodiment) cluster centers to initialize each cluster.

[0061] 2. Traverse all data points and assign each data point to the cluster represented by the cluster center closest to it.

[0062] 3. Recalculate the cluster center of each cluster.

[0063] 4. Repeat steps 2 to 3 until convergence is reached.

[0064] The risk prediction model is optimized using the optimization module to obtain the final model.

[0065] In response to the dynamic changes in transaction patterns and fraud methods, new fraud transaction samples are regularly extracted from real-time data streams, and existing risk rules and models are continuously optimized and updated to improve the accuracy and coverage of fraud identification.

[0066] According to the dynamic changes of transaction patterns and fraud methods, real-time transaction data streams are obtained, and new fraud transaction samples are extracted from them; the extracted new fraud transaction samples are analyzed to determine whether they comply with the existing risk rules. If not, new risk rules are generated based on the sample characteristics and added to the risk rule library; by adding new fraud transaction samples to the training set, the existing fraud identification model is incrementally trained and optimized so that the model can adapt to new fraud patterns.

[0067] For example, a new fraudulent transaction sample shows that the fraudster used the same IP address to conduct multiple large transactions in a short period of time, and the recipients of these transactions were all newly registered accounts. This may indicate that this is a new type of financial fraud. Generate new risk rules based on sample characteristics. For example, based on the above analysis, a new risk rule can be generated: "If the same IP address initiates more than 5 transactions exceeding 10,000 yuan within 1 hour, and the recipient account registration time is less than 7 days, then the transaction is high risk." This rule is added to the risk rule library for real-time monitoring.

[0068] Finally, the prediction module completes the risk prediction based on the final model.

[0069] It should be noted that the above-mentioned financial transaction fraud risk prediction system is embodied in the form of functional units. The term "module" here can be implemented in the form of software and / or hardware, and is not specifically limited to this.

[0070] For example, a "module" may be a software program, a hardware circuit, or a combination of the two that implements the above functions. The hardware circuit may include an application specific integrated circuit (ASIC), an electronic circuit, a processor (such as a shared processor, a dedicated processor, or a group processor, etc.) and a memory for executing one or more software or firmware programs, a combined logic circuit, and / or other suitable components that support the described functions.

[0071] The embodiments of the present disclosure are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present disclosure should be included in the scope of protection of the present disclosure.

Claims

1. A method for predicting financial transaction fraud risk, characterized in that: The method comprises: Collect financial transaction data and pre-process it to obtain processed data; Based on the processed data, construct a risk prediction model; Optimizing the risk prediction model to obtain a final model; Based on the final model, risk prediction is completed.

2. The method for predicting financial transaction fraud risk according to claim 1, characterized in that: The ETL tool is used to clean, transform and load the acquired financial transaction data to obtain the processed data; the method includes: performing quality inspection on the acquired financial transaction data according to pre-established data quality rules to form a data quality report; and using a data cleaning method to process the quality problem data identified in the data quality report.

3. The method for predicting financial transaction fraud risk according to claim 1, characterized in that: Based on the processed data, a fraud risk knowledge graph is constructed using graph database technology, and a global fraud risk view is formed by mining the complex relationship network between transaction entities; When the identified association relationship meets the preset fraud risk judgment rules, the transaction entity is judged to have a fraud risk.

4. The method for predicting financial transaction fraud risk according to claim 3, characterized in that: According to the processed data, based on the constructed risk knowledge graph, a random walk algorithm is used to calculate the fraud risk score of each transaction entity; and transaction entities with similar fraud risk scores are clustered through a clustering algorithm to obtain a group of transaction entities with similar fraud risks.

5. The method for predicting financial transaction fraud risk according to claim 1, characterized in that: Based on the processed data, in response to the dynamic changes in transaction patterns and fraud methods, new fraud transaction samples are regularly extracted from real-time data streams, and the existing risk rules and risk prediction models are continuously optimized and updated to improve the accuracy and coverage of fraud identification, thereby obtaining the final model.

6. A financial transaction fraud risk prediction system, the system being used to implement the method according to any one of claims 1 to 5, characterized in that: include: Acquisition module, construction module, optimization module and prediction module; The acquisition module is used to collect financial transaction data and perform pre-processing to obtain processed data; The construction module is used to construct a risk prediction model based on the processed data; The optimization module is used to optimize the risk prediction model to obtain a final model; The prediction module is used to complete risk prediction based on the final model.

7. The financial transaction fraud risk prediction system according to claim 6, characterized in that: The acquisition module uses ETL tools to clean, convert and load the acquired financial transaction data to obtain the processed data; the method includes: performing quality inspection on the acquired financial transaction data according to pre-established data quality rules to form a data quality report; and using a data cleaning method to process the quality problem data identified in the data quality report.

8. The financial transaction fraud risk prediction system according to claim 6, characterized in that: The construction module constructs a fraud risk knowledge graph based on the processed data using graph database technology, and forms a global fraud risk view by mining the complex relationship network between transaction entities; When the identified association relationship meets the preset fraud risk judgment rules, the transaction entity is judged to have a fraud risk.

Images