Financial service business management system and method based on knowledge graph
Through the financial service business management method based on the knowledge graph, the user's financial service chain and knowledge graph are constructed, and the abnormal risk weight and correlation are calculated, the problem of insufficient analysis accuracy in the existing technology is solved, and more accurate risk identification and management is achieved.
Patent Information
- Application Number
- CN202510312670.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-17
- Publication Date
- 2025-07-04
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing financial service business management technology has problems such as insufficient analysis accuracy and late risk warning in terms of abnormal risk identification and user association analysis, and it is difficult to capture the user's multi-level business behavior characteristics and the comprehensive risk assessment capabilities across business types and transaction amount ranges.
Based on the knowledge graph, the user's financial service chain is constructed, business type, transaction amount and abnormal marking nodes are extracted, three-dimensional risk matrix and two-dimensional image matrix are constructed, abnormal risk weight and correlation degree are calculated, and structured expression and dynamic risk assessment are realized through multi-level risk quantification and user correlation analysis.
It improves the accuracy and efficiency of risk identification, can promptly warn of potential risks, improves the intelligence level of financial service business management, and reduces business risks.
Smart Images
Figure CN120258981A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of service business management, and specifically provides a financial service business management system and method based on a knowledge graph. Background Art
[0002] In recent years, the financial service industry has achieved rapid development under the impetus of digital transformation, gradually extending from traditional offline network services to online services. Intelligent and automated management has become the key direction for improving service efficiency and risk control capabilities. With the in-depth application of technologies such as big data, artificial intelligence, and blockchain, financial institutions can more accurately identify customer needs, optimize business processes, and achieve differentiated services through data-driven decision-making. In this context, as a structured knowledge representation form, the knowledge graph has gradually become an important means to improve the refined level of business management in the financial service field due to its outstanding advantages in multi-source heterogeneous data integration, entity relationship reasoning, and intelligent decision support. However, these technologies mainly focus on single-dimensional analysis, such as the risk judgment of individual transactions, and cannot comprehensively reveal the multi-dimensional correlation characteristics in complex financial behaviors. Especially in the detection and correlation analysis of user abnormal behaviors, there are problems such as insufficient analysis accuracy and lagging risk warnings.
[0003] Existing financial service business management technologies still have many deficiencies in abnormal risk identification and user correlation analysis. On the one hand, traditional single-dimensional data analysis methods are difficult to capture the multi-level business behavior characteristics of users and lack the comprehensive risk assessment ability across business types and transaction amount ranges, resulting in the limitation of abnormal risk identification to isolated business scenarios and making it difficult to effectively capture potential risk chains. On the other hand, the risk correlation analysis between users generally relies on static data comparison and fails to consider the risk linkage effect brought by dynamic business changes, resulting in low warning accuracy and insufficient risk diffusion prevention and control capabilities. Summary of the Invention
[0004] The purpose of the present invention is to provide a financial service business management system and method based on a knowledge graph to solve the problems raised in the above background art.
[0005] To solve the above technical problems, the present invention provides the following technical solutions:
[0006] A financial service business management method based on a knowledge graph, the method comprising the following steps: Step S1: After obtaining user authorization, obtain the online financial service logs of the user in the financial service business management platform, and construct a user financial service chain; Step S2: Based on the user financial service chain, construct a financial service knowledge graph for a single user; Step S3: Based on the financial service knowledge graph, extract business type nodes, transaction amount nodes, and anomaly flag nodes; construct a three-dimensional risk matrix, and calculate the anomaly risk weight of a single business type of a single user in a single transaction amount range; Step S4: Based on the anomaly risk weight, construct a two-dimensional image matrix for a single user, calculate the horizontal gradient and the vertical gradient, calculate the comprehensive linkage gradient of the two-dimensional image matrix of the single user, calculate the correlation degree between users, preset a threshold, and analyze and perform financial service business management.
[0007] As a preferred solution of the financial service business management method based on a knowledge graph of the present invention, after obtaining user authorization, obtain the online financial service logs of the user in the financial service business management platform, the financial service logs including the business type, business status, and transaction amount when the user handles financial service business, the business status including a main status and a status attribute, the main status including in progress, completed, and terminated, and the status attribute including a timestamp (the specific time when the status occurs, such as the approval time, the loan disbursement time) and an anomaly flag (whether it is an abnormal status, such as failure, rejection, and record the reason, such as "insufficient credit").
[0008] Based on the financial service logs, construct a user financial service chain as follows:
[0009] Set user financial service chain nodes, the user financial service chain nodes including business type nodes, main status nodes, timestamp nodes, anomaly flag nodes, and transaction amount nodes.
[0010] Taking the timestamp as the unit, construct an independent user financial service chain for each timestamp, the user financial service chain being sequentially connected through business type nodes, main status nodes, transaction amount nodes, anomaly flag nodes, and timestamp nodes.
[0011] As a preferred solution of the financial service business management method based on a knowledge graph of the present invention, based on the user financial service chain, construct a financial service knowledge graph for the i-th user as follows:
[0012] Obtain the user ID of the i-th user, and use the user ID as the starting entity of the financial service knowledge graph; use the user financial service chain as the edge of the financial service knowledge graph; use the business type as the ending entity of the financial service knowledge graph.
[0013] Construct a financial service knowledge graph for the i-th user based on the connection relationship of start entity-edge-end entity, obtain the user's financial service chain for the i-th user at all timestamps, and add it to the financial service knowledge graph in sequence according to the timestamp. Moreover, if the business types of the i-th user are the same at different timestamps, add an edge between the corresponding start entity and end entity.
[0014] As a preferred solution of the financial service business management method based on a knowledge graph according to the present invention, based on the financial service knowledge graph, extract all business type nodes of the i-th user, as well as all transaction amount nodes and anomaly flag nodes corresponding to a single business type node.
[0015] Evenly divide the all transaction amount nodes into N transaction amount intervals (such as 0-10,000, 10,000-100,000, above 100,000), and count the anomaly flag distribution of each business type in each transaction amount interval.
[0016] Construct a three-dimensional risk matrix, where the row dimension, column dimension, and layer dimension of the three-dimensional risk matrix are business type, transaction amount interval, and anomaly flag respectively; in the three-dimensional risk matrix, a cell represents a risk probability, and record the risk probability caused by the k-th anomaly flag for the a-th business type of the i-th user in the n-th transaction amount interval as R a,n,k (i), and the calculation formula of the risk probability R a,n,k (i) is: where N a,n,k represents the occurrence times of the k-th anomaly flag for the a-th business type in the n-th transaction amount interval, and N a,n represents the total transaction times of the a-th business type in the n-th transaction amount interval.
[0017] According to the a-th business type and the n-th transaction amount interval of the i-th user, lock the corresponding row dimension and column dimension in the three-dimensional risk matrix. According to the locked row dimension and column dimension, extract the risk probabilities under all anomaly flags in the corresponding layer dimension, and calculate the anomaly risk weight of the a-th business type of the i-th user in the n-th transaction amount interval. The calculation formula is as follows:
[0018]
[0019] where RTR a,n (i) represents the anomaly risk weight of the a-th business type of the i-th user in the n-th transaction amount interval, and M represents all anomaly flags in the corresponding layer dimension.
[0020] As a preferred solution of the financial service business management method based on the knowledge graph of the present invention, based on the abnormal risk weight RTR a,n (i) of the a-th business type of the i-th user in the n-th transaction amount range, construct a two-dimensional image matrix of the i-th user, specifically as follows:
[0021]
[0022] Among them, RTR A,N (i) represents the abnormal risk weight of the A-th business type of the i-th user in the N-th transaction amount range, and A represents the total number of business types of the i-th user.
[0023] In the present invention, the elements in the two-dimensional image matrix can be approximately regarded as the pixel grids in the corresponding grayscale image converted from the three-dimensional risk matrix. The business type and the transaction amount range can be approximately represented as the length and width of the pixel grid, and the abnormal risk weight can be approximately represented as the depth of the pixel grid.
[0024] Calculate the horizontal gradient and vertical gradient of the two-dimensional image matrix, and the calculation formulas are as follows:
[0025]
[0026]
[0027] Among them, HG i (a,n) represents the horizontal gradient of the two-dimensional image matrix of the i-th user, and LG i (a,n) represents the vertical gradient of the two-dimensional image matrix of the i-th user, RTR a,n+1 (i) represents the abnormal risk weight of the a-th business type of the i-th user in the n+1-th transaction amount range, and RTR a+1,n (i) represents the abnormal risk weight of the a+1-th business type of the i-th user in the n-th transaction amount range.
[0028] Based on the horizontal gradient HG i (a,n) and the vertical gradient LG i (a,n), calculate the comprehensive linkage gradient of the two-dimensional image matrix of the i-th user, and the calculation formula is: Among them, CG i (a,n) represents the comprehensive linkage gradient of the two-dimensional image matrix of the i-th user.
[0029] Based on the comprehensive linkage gradient CG i (a,n) of the two-dimensional image matrix of the i-th user, calculate the correlation degree between the i-th user and the i+1-th user, and the calculation formula is: CD i→i+1 =|CGi (a,n)-CG i+1 (a,n)|, where CD i→i+1 represents the degree of association between the i-th user and the (i + 1)-th user, and CG i+1 (a,n) represents the comprehensive linkage gradient of the two-dimensional image matrix of the (i + 1)-th user.
[0030] A preset degree-of-association threshold. If the degree of association CD between the i-th user and the (i + 1)-th user i→i+1 is less than the degree-of-association threshold, it is determined that the i-th user and the (i + 1)-th user are similar.
[0031] If the i-th user has no anomaly mark when initiating a new service (service application is successful), while the (i + 1)-th user has an anomaly mark when initiating a new service (service application is unsuccessful), then a warning is issued to the business approval staff of the (i + 1)-th user.
[0032] A financial service business management system based on a knowledge graph. This system includes: a data acquisition and service chain construction module, a financial service knowledge graph construction module, a matrix construction and weight calculation module, and a gradient calculation and analysis module.
[0033] The data acquisition and service chain construction module: After obtaining user authorization, it acquires the online financial service logs of users in the financial service business management platform and constructs a user financial service chain.
[0034] The financial service knowledge graph construction module: Based on the user financial service chain, it constructs a financial service knowledge graph for a single user.
[0035] The matrix construction and weight calculation module: Based on the financial service knowledge graph, it extracts business type nodes, transaction amount nodes, and anomaly mark nodes; constructs a three-dimensional risk matrix, and calculates the anomaly risk weight of a single business type of a single user in a single transaction amount range.
[0036] The gradient calculation and analysis module: Based on the anomaly risk weight, it constructs a two-dimensional image matrix of a single user, calculates the horizontal gradient and the vertical gradient, calculates the comprehensive linkage gradient of the two-dimensional image matrix of a single user, calculates the degree of association between users, sets a preset threshold, and analyzes and conducts financial service business management.
[0037] Further, the data acquisition and service chain construction module includes a data acquisition unit and a service chain construction unit.
[0038] The data acquisition unit: After being authorized by the user, it acquires the online financial service logs of the user in the financial service business management platform. The online financial service logs include the business type, business status, and transaction amount when the user handles financial service business. The business status includes a main status and a status attribute. The main status includes in progress, completed, and terminated. The status attribute includes a timestamp and an exception flag.
[0039] The service chain construction unit: Based on the online financial service logs, it constructs the user's financial service chain as follows: Set the user's financial service chain nodes. The user's financial service chain nodes include a business type node, a main status node, a timestamp node, an exception flag node, and a transaction amount node. Taking the timestamp as the unit, it constructs an independent user financial service chain for each timestamp. The user financial service chain is sequentially connected through the business type node, the main status node, the transaction amount node, the exception flag node, and the timestamp node.
[0040] Further, the financial service knowledge graph construction module includes a knowledge graph construction unit.
[0041] The knowledge graph construction unit: Based on the user's financial service chain, it constructs the financial service knowledge graph of the i-th user as follows: Obtain the user ID of a single user and use the user ID as the starting entity of the financial service knowledge graph. Use the user's financial service chain as the edge of the financial service knowledge graph. Use the business type as the ending entity of the financial service knowledge graph. With the connection relationship of starting entity - edge - ending entity, it constructs the financial service knowledge graph of a single user, obtains the user's financial service chain at all timestamps of a single user, and adds them to the financial service knowledge graph in sequence according to the timestamp. And if the business types of a single user at different timestamps are the same, an additional edge is added between the corresponding starting entity and ending entity.
[0042] Further, the matrix construction and weight calculation module includes a matrix construction unit and a weight calculation unit.
[0043] The matrix construction unit: Based on the financial service knowledge graph, it extracts all business type nodes of a single user, as well as all transaction amount nodes and exception flag nodes corresponding to a single business type node.
[0044] For all the transaction amount nodes, they are evenly divided into N transaction amount intervals, and the distribution of exception flags in each transaction amount interval for each business type is counted.
[0045] Construct a three-dimensional risk matrix, where the row dimension, column dimension, and layer dimension of the three-dimensional risk matrix are business type, transaction amount range, and anomaly flag respectively; in the three-dimensional risk matrix, a cell represents a risk probability, and the risk probability is caused by a single anomaly flag for a single business type of a single user in a single transaction amount range.
[0046] The weight calculation unit: According to the single business type and single transaction amount range of a single user, lock the corresponding row dimension and column dimension in the three-dimensional risk matrix. According to the locked row dimension and column dimension, extract the risk probabilities under all anomaly flags in the corresponding layer dimension, and calculate the anomaly risk weight of the single business type of the single user in the single transaction amount range.
[0047] Further, the gradient calculation and analysis module includes a gradient calculation unit and an analysis unit.
[0048] The gradient calculation unit: Based on the anomaly risk weight of a single business type of a single user in a single transaction amount range, construct a two-dimensional image matrix of the single user; calculate the horizontal gradient and vertical gradient of the two-dimensional image matrix.
[0049] The analysis unit: Based on the horizontal gradient and vertical gradient, calculate the comprehensive linkage gradient of the two-dimensional image matrix of the single user; based on the comprehensive linkage gradient of the two-dimensional image matrix of the single user, calculate the correlation degree between the single user and the next user; preset a correlation degree threshold. If the correlation degree between the single user and the next user is less than the correlation degree threshold, then determine that the single user is similar to the next user; if there is no anomaly flag when the single user initiates a new business, and there is an anomaly flag when the next user initiates a new business, then issue a warning to the business approval staff of the next user.
[0050] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: In a financial service business management system and method based on a knowledge graph provided by the present invention, by constructing a user financial service chain, a data foundation is laid for the subsequent construction of the knowledge graph. This chain details the business type, status, transaction amount, and anomaly marks, ensuring the integrity and accuracy of the data; taking the user ID as the starting entity, the financial service chain as the edge, and the business type as the ending entity, a user-specific financial service knowledge graph is constructed, realizing a structured expression of the user's financial behavior, which is convenient for further mining and analysis; extracting key nodes from the knowledge graph, constructing a three-dimensional risk matrix, associating the business type, transaction amount range, and anomaly marks, quantitatively calculating the risk probability in a refined manner, and obtaining the anomaly risk weight, which can more accurately capture potential risks; generating a two-dimensional image matrix based on the anomaly risk weight, calculating the horizontal and vertical gradients and the comprehensive linkage gradient, further revealing the correlation between the user's transaction behaviors, and by setting a correlation threshold, distinguishing the similarity between users, and then when abnormal user behaviors occur, timely issuing a warning to the business approval personnel; overall, this method not only realizes the structured expression of the user's financial service behavior, but also improves the intelligent level of financial service business management through multi-level risk quantification and user correlation analysis, effectively reducing the business risk, improving the service efficiency, and providing a more accurate and efficient risk identification and management means for financial institutions. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification, and are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention.
[0052] Figure 1 is a schematic diagram of the steps of a financial service business management method based on a knowledge graph of the present invention;
[0053] Figure 2 is a schematic diagram of the structure of a financial service business management system based on a knowledge graph of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.
[0055] Please refer to Figure 1 , in the first embodiment: A financial service business management method based on a knowledge graph is provided, and this method includes the following steps:
[0056] Step S1: After obtaining user authorization, retrieve the online financial service logs of the user in the financial service business management platform and construct a user financial service chain.
[0057] Specifically, after obtaining user authorization, retrieve the online financial service logs of the user in the financial service business management platform. The online financial service logs include the business type, business status, and transaction amount when the user handles financial service business. The business status includes a main status and status attributes. The main status includes in progress, completed, and terminated. The status attributes include a timestamp (the specific time when the status occurs (such as the approval time, loan disbursement time)) and an exception flag (whether it is an abnormal status (such as failure, rejection), and record the reason (such as "insufficient credit"));
[0058] Furthermore, based on the online financial service logs, construct a user financial service chain as follows:
[0059] Set user financial service chain nodes. The user financial service chain nodes include business type nodes, main status nodes, timestamp nodes, exception flag nodes, and transaction amount nodes;
[0060] Taking the timestamp as the unit, construct an independent user financial service chain for each timestamp. The user financial service chain is sequentially connected through business type nodes, main status nodes, transaction amount nodes, exception flag nodes, and timestamp nodes.
[0061] It should be noted that in this step, by retrieving the online financial service logs authorized by the user, extracting key data such as the user's business type, business status, timestamp, and exception flag, and constructing a user financial service chain, it realizes the comprehensive capture and structured arrangement of the user's full-process behavior; its role is to form a time-sequential user financial behavior chain, which can not only completely reflect the user's historical operation trajectory but also lay a foundation for the subsequent construction of the knowledge graph; it can improve the granularity and accuracy of the data, make the dynamic evolution of the user's behavior visible, and provide high-quality data support for risk identification and business management.
[0062] Step S2: Based on the user financial service chain, construct a financial service knowledge graph for a single user.
[0063] Specifically, based on the user financial service chain, construct a financial service knowledge graph for the i-th user as follows:
[0064] Obtain the user ID of the i-th user and use the user ID as the starting entity of the financial service knowledge graph; use the user financial service chain as the edge of the financial service knowledge graph; use the business type as the ending entity of the financial service knowledge graph;
[0065] Construct the financial service knowledge graph of the i-th user based on the connection relationship of the starting entity-edge-ending entity, obtain the user's financial service chain at all timestamps of the i-th user, and add it to the financial service knowledge graph in sequence according to the timestamp. Moreover, if the business types of the i-th user are the same at different timestamps, an edge is added between the corresponding starting entity and ending entity.
[0066] It should be noted that this step constructs the user's financial service knowledge graph based on the user's financial service chain. Through the structured relationship of "user ID-edge-business type", the multiple transaction behaviors of the user are associated, realizing the leap from a linear chain to a network graph of user behavior data; the role of this step is to transform the discrete behavior nodes of the user into an analyzable graph network, enhancing the data linkage and traceability; realizing the multi-dimensional and multi-level associated expression of the user's financial behavior, providing a richer data context for the mining of complex risk patterns and anomaly identification.
[0067] Step S3: Based on the financial service knowledge graph, extract the business type nodes, transaction amount nodes, and anomaly marker nodes; construct a three-dimensional risk matrix, and calculate the anomaly risk weight of a single business type of a single user in a single transaction amount interval.
[0068] Specifically, based on the financial service knowledge graph, extract all the business type nodes of the i-th user, as well as all the transaction amount nodes and anomaly marker nodes corresponding to a single business type node;
[0069] For all the transaction amount nodes, evenly divide them into N transaction amount intervals ((such as 0-10,000, 10,000-100,000, above 100,000)), and count the distribution of anomaly markers for each business type in each transaction amount interval;
[0070] Furthermore, construct a three-dimensional risk matrix. The row dimension, column dimension, and layer dimension of the three-dimensional risk matrix are business type, transaction amount interval, and anomaly marker respectively; in the three-dimensional risk matrix, a cell represents a risk probability. Denote the risk probability caused by the k-th anomaly marker for the a-th business type of the i-th user in the n-th transaction amount interval as R a,n,k (i), and the calculation formula for the risk probability R a,n,k (i) is: Among them, N a,n,k represents the number of occurrences of the k-th anomaly marker for the a-th business type in the n-th transaction amount interval, and N a,n represents the total number of transactions for the a-th business type in the n-th transaction amount interval;
[0071] Lock the corresponding row dimension and column dimension in the three-dimensional risk matrix according to the a-th business type and the n-th transaction amount range of the i-th user. According to the locked row dimension and column dimension, extract the risk probabilities under all abnormal marks in the corresponding layer dimension, and calculate the abnormal risk weight of the a-th business type of the i-th user in the n-th transaction amount range. The calculation formula is as follows:
[0072]
[0073] Among them, RTR a,n (i) represents the abnormal risk weight of the a-th business type of the i-th user in the n-th transaction amount range, and M represents all abnormal marks in the corresponding layer dimension.
[0074] It should be noted that in this step, by extracting business type nodes, transaction amount nodes, and abnormal mark nodes, a three-dimensional risk matrix is constructed, the risk probability is divided according to the transaction amount range, and the abnormal risk weight is calculated, realizing the accurate quantification of the risk of a single user's business in a specific transaction amount range. Its role is to refine the risk analysis dimension, from a simple abnormal mark judgment to a comprehensive evaluation model with a three-layer structure of "business type - transaction amount - abnormal mark"; it can identify the potential risk patterns of a single user under different business types and transaction ranges, improve the granularity and accuracy of risk judgment, and facilitate the subsequent implementation of targeted business management and intervention measures.
[0075] Step S4: Based on the abnormal risk weight, construct a two-dimensional image matrix of a single user, calculate the horizontal gradient and vertical gradient, calculate the comprehensive linkage gradient of the two-dimensional image matrix of a single user, calculate the correlation degree between users, set a preset threshold, and analyze and conduct financial service business management.
[0076] Specifically, based on the abnormal risk weight RTR a,m (i) of the a-th business type of the i-th user in the n-th transaction amount range, construct the two-dimensional image matrix of the i-th user, as follows:
[0077]
[0078] Among them, RTR A,N (i) represents the abnormal risk weight of the A-th business type of the i-th user in the N-th transaction amount range, and A represents the total number of business types of the i-th user;
[0079] Furthermore, calculate the horizontal gradient and vertical gradient of the two-dimensional image matrix. The calculation formula is as follows:
[0080]
[0081] Among them, HGi (a,n) represents the horizontal gradient of the two-dimensional image matrix of the i-th user, HG i (a,n) represents the vertical gradient of the two-dimensional image matrix of the i-th user, RTR a,n+1 (i) represents the abnormal risk weight of the a-th business type of the i-th user in the (n + 1)-th transaction amount interval, RTR a+1,n (i) represents the abnormal risk weight of the (a + 1)-th business type of the i-th user in the n-th transaction amount interval;
[0082] Based on the horizontal gradient HG i (a,n) and the vertical gradient LG i (a,n), calculate the comprehensive linkage gradient of the two-dimensional image matrix of the i-th user. The calculation formula is: where CG i (a,n) represents the comprehensive linkage gradient of the two-dimensional image matrix of the i-th user;
[0083] Furthermore, based on the comprehensive linkage gradient CG i (a,n) of the two-dimensional image matrix of the i-th user, calculate the correlation degree between the i-th user and the (i + 1)-th user. The calculation formula is: CD i→i+1 = |CG i (a,n) - CG i+1 (a,n)|, where CD i→i+1 represents the correlation degree between the i-th user and the (i + 1)-th user, and CG i+1 (a,n) represents the comprehensive linkage gradient of the two-dimensional image matrix of the (i + 1)-th user;
[0084] Preset a correlation degree threshold. If the correlation degree CD i→i+1 between the i-th user and the (i + 1)-th user is less than the correlation degree threshold, it is determined that the i-th user and the (i + 1)-th user are similar;
[0085] If there is no abnormal mark (business application is successful) when the i-th user initiates a new business, while there is an abnormal mark (business application is unsuccessful) when the (i + 1)-th user initiates a new business, a warning is issued to the business approval staff of the (i + 1)-th user.
[0086] It should be noted that in this step, a two-dimensional image matrix is constructed based on the abnormal risk weight, the horizontal gradient, the vertical gradient and the comprehensive linkage gradient are calculated, the correlation degree between users is further calculated, and the similarity between users is determined through the correlation degree threshold, and finally the abnormal warning trigger mechanism is realized. Its function is to convert the financial behavior of users into a quantifiable image matrix, and use the change of the image gradient to reflect the comprehensive change trend of users' behavior, breaking the limitation that traditional risk analysis can only be based on single-point data. It can realize dynamic risk assessment across users and across businesses, and can trigger the warning mechanism for potential risk users in a timely manner when detecting abnormal behaviors of similar users, thereby improving the risk prevention and control ability of the financial service platform and reducing potential losses.
[0087] Please refer to Figure 2 , in the second embodiment: A financial service business management system based on a knowledge graph is provided. The system includes: a data acquisition and service chain construction module, a financial service knowledge graph construction module, a matrix construction and weight calculation module, and a gradient calculation and analysis module.
[0088] The data acquisition and service chain construction module: After obtaining user authorization, it acquires the online financial service logs of users in the financial service business management platform and constructs a user financial service chain.
[0089] The financial service knowledge graph construction module: Based on the user financial service chain, it constructs a financial service knowledge graph of a single user.
[0090] The matrix construction and weight calculation module: Based on the financial service knowledge graph, it extracts business type nodes, transaction amount nodes and abnormal mark nodes; constructs a three-dimensional risk matrix, and calculates the abnormal risk weight of a single business type of a single user in a single transaction amount range.
[0091] The gradient calculation and analysis module: Based on the abnormal risk weight, it constructs a two-dimensional image matrix of a single user, calculates the horizontal gradient and the vertical gradient, calculates the comprehensive linkage gradient of the two-dimensional image matrix of a single user, calculates the correlation degree between users, sets a preset threshold, and analyzes and manages the financial service business.
[0092] Furthermore, the data acquisition and service chain construction module includes a data acquisition unit and a service chain construction unit.
[0093] The data acquisition unit: After obtaining user authorization, it acquires the online financial service logs of users in the financial service business management platform. The online financial service logs include the business type, business status and transaction amount when users handle financial service business. The business status includes a main status and a status attribute. The main status includes in progress, completed and terminated. The status attribute includes a time stamp and an abnormal mark.
[0094] The service chain construction unit: Based on the online financial service logs, construct the user's financial service chain as follows: Set the nodes of the user's financial service chain, where the nodes of the user's financial service chain include business type nodes, main status nodes, timestamp nodes, exception flag nodes, and transaction amount nodes; Taking the timestamp as the unit, construct an independent user financial service chain for each timestamp, and the user financial service chain is sequentially connected through business type nodes, main status nodes, transaction amount nodes, exception flag nodes, and timestamp nodes.
[0095] Furthermore, the financial service knowledge graph construction module includes a knowledge graph construction unit.
[0096] The knowledge graph construction unit: Based on the user's financial service chain, construct the financial service knowledge graph of the i-th user as follows: Obtain the user ID of a single user, and use the user ID as the starting entity of the financial service knowledge graph; Use the user's financial service chain as the edge of the financial service knowledge graph; Use the business type as the ending entity of the financial service knowledge graph; Construct the financial service knowledge graph of a single user in the connection relationship of starting entity - edge - ending entity, obtain the user's financial service chain of a single user at all timestamps, and add them to the financial service knowledge graph in sequence according to the timestamp. Moreover, if the business types of a single user at different timestamps are the same, add an edge between the corresponding starting entity and ending entity.
[0097] Furthermore, the matrix construction and weight calculation module includes a matrix construction unit and a weight calculation unit.
[0098] The matrix construction unit: Based on the financial service knowledge graph, extract all business type nodes of a single user, as well as all transaction amount nodes and exception flag nodes corresponding to a single business type node.
[0099] For all the transaction amount nodes, evenly divide them into N transaction amount intervals, and count the distribution of exception flags for each business type in each transaction amount interval.
[0100] Construct a three-dimensional risk matrix, where the row dimension, column dimension, and layer dimension of the three-dimensional risk matrix are business type, transaction amount interval, and exception flag respectively; In the three-dimensional risk matrix, a cell represents a risk probability, and the risk probability is caused by a single exception flag of a single business type of a single user in a single transaction amount interval.
[0101] The weight calculation unit: According to the single business type and single transaction amount range of a single user, lock the corresponding row dimension and column dimension in the three-dimensional risk matrix. According to the locked row dimension and column dimension, extract the risk probabilities under all abnormal marks in the corresponding layer dimension, and calculate the abnormal risk weight of the single business type of a single user in the single transaction amount range.
[0102] Further, the gradient calculation and analysis module includes a gradient calculation unit and an analysis unit.
[0103] The gradient calculation unit: Based on the abnormal risk weight of the single business type of a single user in the single transaction amount range, construct a two-dimensional image matrix of the single user; calculate the horizontal gradient and vertical gradient of the two-dimensional image matrix.
[0104] The analysis unit: Based on the horizontal gradient and vertical gradient, calculate the comprehensive linkage gradient of the two-dimensional image matrix of the single user; based on the comprehensive linkage gradient of the two-dimensional image matrix of the single user, calculate the correlation degree between the single user and the next user; preset a correlation degree threshold. If the correlation degree between the single user and the next user is less than the correlation degree threshold, it is determined that the single user is similar to the next user; if there is no abnormal mark when the single user initiates a new business, and there is an abnormal mark when the next user initiates a new business, a warning is issued to the business approval staff of the next user.
[0105] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.
[0106] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
Claims
1. A financial service business management method based on a knowledge graph, characterized in that, The method includes the following steps: Step S1: After obtaining user authorization, obtain the online financial service logs of the user in the financial service business management platform, and construct a user financial service chain; Step S2: Based on the user financial service chain, construct a financial service knowledge graph for a single user; Step S3: Based on the financial service knowledge graph, extract business type nodes, transaction amount nodes, and anomaly flag nodes; construct a three-dimensional risk matrix, and calculate the anomaly risk weight of a single business type of a single user in a single transaction amount range; Step S4: Based on the anomaly risk weight, construct a two-dimensional image matrix for a single user, calculate the horizontal gradient and vertical gradient, calculate the comprehensive linkage gradient of the two-dimensional image matrix of a single user, calculate the correlation degree between users, preset a threshold, and analyze and conduct financial service business management.
2. The method for managing financial service operations based on a knowledge graph according to claim 1, wherein, The specific implementation process of step S1 includes: After obtaining user authorization, obtain the online financial service logs of the user in the financial service business management platform. The financial service online logs include the business type, business status, and transaction amount when the user handles financial service business. The business status includes a main status and a status attribute. The main status includes in progress, completed, and terminated. The status attribute includes a timestamp and an anomaly flag; Based on the financial service online logs, construct a user financial service chain as follows: Set user financial service chain nodes. The user financial service chain nodes include business type nodes, main status nodes, timestamp nodes, anomaly flag nodes, and transaction amount nodes; Taking the timestamp as the unit, construct an independent user financial service chain for each timestamp. The user financial service chain is sequentially connected by business type nodes, main status nodes, transaction amount nodes, anomaly flag nodes, and timestamp nodes.
3. The method for managing financial service operations based on a knowledge graph according to claim 2, wherein, The specific implementation process of step S2 includes: Based on the user financial service chain, construct a financial service knowledge graph for the i-th user as follows: Obtain the user ID of the i-th user, and use the user ID as the starting entity of the financial service knowledge graph; use the user financial service chain as the edge of the financial service knowledge graph; use the business type as the ending entity of the financial service knowledge graph; Construct a financial service knowledge graph for the i-th user with the connection relationship of starting entity - edge - ending entity, obtain the user financial service chain of the i-th user at all timestamps, and add them to the financial service knowledge graph in sequence according to the timestamp. And if the business types of the i-th user are the same at different timestamps, add an edge between the corresponding starting entity and ending entity.
4. A method for managing financial service operations based on a knowledge graph according to claim 3, characterized in that, The specific implementation process of step S3 includes: Based on the financial service knowledge graph, extract all business type nodes of the i-th user, as well as all transaction amount nodes and anomaly flag nodes corresponding to a single business type node; For all the transaction amount nodes, evenly divide them into N transaction amount ranges, and count the anomaly flag distribution of each business type in each transaction amount range; Construct a three-dimensional risk matrix, where the row dimension, column dimension, and layer dimension of the three-dimensional risk matrix are business type, transaction amount range, and anomaly flag respectively; in the three-dimensional risk matrix, a cell represents a risk probability. Denote the risk probability caused by the k-th anomaly flag for the a-th business type of the i-th user in the n-th transaction amount range as R a,n,k (i), and the risk probability R a,n,k (i) is calculated by the formula: where N a,n,k represents the number of occurrences of the k-th anomaly flag for the a-th business type in the n-th transaction amount range, and N a,n represents the total number of transactions for the a-th business type in the n-th transaction amount range; Lock the corresponding row dimension and column dimension in the three-dimensional risk matrix according to the a-th business type and the n-th transaction amount range of the i-th user. According to the locked row dimension and column dimension, extract the risk probabilities under all abnormal marks in the corresponding layer dimension, and calculate the abnormal risk weight of the a-th business type of the i-th user in the n-th transaction amount range. The calculation formula is as follows: Among them, RTR a,n (i) represents the abnormal risk weight of the a-th service type of the i-th user under the n-th transaction amount range, and M represents all abnormal marks in the corresponding layer dimension.
5. A method for managing financial service operations based on a knowledge graph according to claim 4, characterized in that, The specific implementation process of step S4 includes: Abnormal risk weight RTR for the a-th business type of the i-th user in the n-th transaction amount range a,n (i), construct a two-dimensional image matrix for the i-th user as follows: Among them, RTR A,N (i) represents the abnormal risk weight of the A-th service type of the i-th user under the N-th transaction amount range, and A represents the total number of service types of the i-th user; Calculate the horizontal gradient and vertical gradient of the two-dimensional image matrix. The calculation formula is as follows: Among them, HG i (a,n) represents the horizontal gradient of the two-dimensional image matrix of the i-th user, LG i (a,n) represents the vertical gradient of the two-dimensional image matrix of the i-th user, RTR a,n+1 (i) represents the abnormal risk weight of the a-th business type of the i-th user in the (n + 1)-th transaction amount interval, RTR a+1,n (i) represents the abnormal risk weight of the (a + 1)-th business type of the i-th user in the n-th transaction amount interval; Based on the horizontal gradient HG i (a, n) and the vertical gradient LG i (a, n), calculate the comprehensive linkage gradient of the 2D image matrix of the i-th user. The calculation formula is: Among them, CG i (a, n) represents the comprehensive linkage gradient of the 2D image matrix of the i-th user; Comprehensive linkage gradient CG based on the two-dimensional image matrix of the i-th user i (a,n), calculate the correlation degree between the i-th user and the (i + 1)-th user, and the calculation formula is: CD i→i+1 = |CG i (a,n) - CG i+1 (a,n)|, where CD i→i+1 represents the correlation degree between the i-th user and the (i + 1)-th user, and CG i+1 (a,n) represents the comprehensive linkage gradient of the two-dimensional image matrix of the (i + 1)-th user; A preset correlation threshold. If the correlation CD between the i-th user and the (i + 1)-th user i→i+1 is less than the correlation threshold, it is determined that the i-th user is similar to the (i + 1)-th user; If there is no abnormal mark when the i-th user initiates a new business, but there is an abnormal mark when the (i + 1)-th user initiates a new business, then issue a warning to the business approval staff of the (i + 1)-th user.
6. A financial service business management system based on a knowledge graph, which executes a financial service business management method based on a knowledge graph as described in any one of claims 1-5, characterized in that The system includes: a data acquisition and service chain construction module, a financial service knowledge graph construction module, a matrix construction and weight calculation module, and a gradient calculation and analysis module; The data acquisition and service chain construction module: After obtaining user authorization, acquire the online financial service logs of the user in the financial service business management platform, and construct a user financial service chain; The financial service knowledge graph construction module: Based on the user financial service chain, construct a financial service knowledge graph of a single user; The matrix construction and weight calculation module: Based on the financial service knowledge graph, extract business type nodes, transaction amount nodes, and abnormal mark nodes; construct a three-dimensional risk matrix, and calculate the abnormal risk weight of a single business type of a single user in a single transaction amount range; The gradient calculation and analysis module: Based on the abnormal risk weight, construct a two-dimensional image matrix of a single user, calculate the horizontal gradient and vertical gradient, calculate the comprehensive linkage gradient of the two-dimensional image matrix of a single user, calculate the correlation degree between users, set a preset threshold, and analyze and manage financial service operations.
7. A financial service business management system based on a knowledge graph according to claim 6, characterized in that: The data acquisition and service chain construction module includes a data acquisition unit and a service chain construction unit; The data acquisition unit: After obtaining user authorization, acquire the online financial service logs of the user in the financial service business management platform. The online financial service logs include the business type, business status, and transaction amount when the user handles financial service operations. The business status includes a main status and a status attribute. The main status includes in progress, completed, and terminated. The status attribute includes a timestamp and an abnormal mark; The service chain construction unit: Based on the online financial service logs, construct a user financial service chain as follows: Set user financial service chain nodes. The user financial service chain nodes include business type nodes, main status nodes, timestamp nodes, abnormal mark nodes, and transaction amount nodes; Taking the timestamp as the unit, construct an independent user financial service chain for each timestamp. The user financial service chain is sequentially connected through business type nodes, main status nodes, transaction amount nodes, abnormal mark nodes, and timestamp nodes.
8. The financial service business management system based on a knowledge graph according to claim 7, wherein: The financial service knowledge graph construction module includes a knowledge graph construction unit; The knowledge graph construction unit: Based on the user's financial service chain, construct the financial service knowledge graph of the i-th user, specifically as follows: Obtain the user ID of a single user, and use the user ID as the starting entity of the financial service knowledge graph; Use the user's financial service chain as the edge of the financial service knowledge graph; Use the business type as the ending entity of the financial service knowledge graph; Construct the financial service knowledge graph of a single user with the connection relationship of starting entity - edge - ending entity, obtain the user's financial service chain at all timestamps of a single user, and add it to the financial service knowledge graph in sequence according to the timestamp. Moreover, if the business types of a single user at different timestamps are the same, add an edge between the corresponding starting entity and ending entity.
9. A financial service business management system based on a knowledge graph according to claim 8, characterized in that: The matrix construction and weight calculation module includes a matrix construction unit and a weight calculation unit; The matrix construction unit: Based on the financial service knowledge graph, extract all business type nodes of a single user, as well as all transaction amount nodes and anomaly flag nodes corresponding to a single business type node; For all the transaction amount nodes, evenly divide them into N transaction amount intervals, and count the anomaly flag distribution of each business type in each transaction amount interval; Construct a three-dimensional risk matrix, where the row dimension, column dimension, and layer dimension of the three-dimensional risk matrix are business type, transaction amount interval, and anomaly flag respectively; In the three-dimensional risk matrix, a cell represents a risk probability, and the risk probability is caused by a single anomaly flag for a single business type of a single user in a single transaction amount interval; The weight calculation unit: According to a single business type and a single transaction amount interval of a single user, lock the corresponding row dimension and column dimension in the three-dimensional risk matrix. According to the locked row dimension and column dimension, extract the risk probabilities under all anomaly flags in the corresponding layer dimension, and calculate the anomaly risk weight of a single business type of a single user in a single transaction amount interval.
10. A financial service business management system based on a knowledge graph according to claim 9, characterized in that: The gradient calculation and analysis module includes a gradient calculation unit and an analysis unit; The gradient calculation unit: Based on the anomaly risk weight of a single business type of a single user in a single transaction amount interval, construct a two-dimensional image matrix of a single user; Calculate the horizontal gradient and vertical gradient of the two-dimensional image matrix; The analysis unit: Based on the horizontal gradient and vertical gradient, calculate the comprehensive linkage gradient of the two-dimensional image matrix of a single user; Based on the comprehensive linkage gradient of the two-dimensional image matrix of a single user, calculate the correlation degree between a single user and the next user; Preset a correlation degree threshold. If the correlation degree between a single user and the next user is less than the correlation degree threshold, then determine that the single user is similar to the next user; If there is no anomaly flag when a single user initiates a new business, while there is an anomaly flag when the next user initiates a new business, then issue a warning to the business approval staff of the next user.