Credit risk customer management method and device based on knowledge graph, equipment and medium

Abstract

This invention relates to the field of data processing technology and discloses a method, apparatus, device, and medium for credit risk customer management based on a knowledge graph. The method includes: acquiring basic customer information and credit information of the customer to be managed; performing risk event analysis on the basic customer information to obtain first risk event feature information; extracting risk information from the credit information to obtain second risk event feature information; performing knowledge fusion processing on the first and second risk event feature information to construct a credit risk knowledge graph; and analyzing and evaluating the credit risk of the customer to be managed based on the credit risk knowledge graph to obtain a target credit risk score and a target management strategy. Using the method provided in this application, existing credit data can be effectively integrated and utilized to improve the accuracy of the credit risk management process.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, apparatus, equipment and medium for credit risk customer management based on knowledge graphs. Background Technology

[0002] In today's complex and ever-changing financial market environment, lending has undoubtedly become one of the core businesses of financial institutions. Faced with ever-expanding financial business and an explosive growth in financial data, traditional methods of processing this massive, multi-source, and heterogeneous financial data may be insufficient to effectively integrate and utilize it for more precise credit risk management. Therefore, how to better and more effectively integrate and utilize financial data to achieve more accurate credit risk management has become a pressing issue that needs to be addressed. Summary of the Invention

[0003] This invention provides a knowledge graph-based credit risk customer management method, apparatus, equipment, and medium to address the technical problem of how to better and more effectively integrate and utilize financial data to achieve more accurate credit risk management.

[0004] Firstly, a knowledge graph-based approach to credit risk customer management is provided, including:

[0005] Obtain basic customer information and customer credit information of customers to be managed;

[0006] Risk event analysis is performed on basic customer information to obtain the characteristic information of the first risk event;

[0007] Risk information is extracted from customer credit information to obtain the characteristic information of the second risk event;

[0008] Knowledge fusion processing is performed on the characteristic information of the first risk event and the characteristic information of the second risk event to construct a credit risk knowledge graph;

[0009] Based on the credit risk knowledge graph, the credit risk of the clients to be managed is analyzed and evaluated to obtain the target credit risk score and target management strategy.

[0010] Secondly, a knowledge graph-based credit risk customer management device is provided, comprising:

[0011] The acquisition module is used to acquire basic customer information and customer credit information of customers to be managed.

[0012] The first processing module is used to perform risk event analysis on the customer's basic information to obtain the first risk event characteristic information;

[0013] The second processing module is used to extract risk information from customer credit information to obtain the second risk event feature information.

[0014] The third processing module is used to perform knowledge fusion processing on the feature information of the first risk event and the feature information of the second risk event in order to construct a credit risk knowledge graph.

[0015] The fourth processing module is used to analyze and evaluate the credit risk of the clients to be managed based on the credit risk knowledge graph, so as to obtain the target credit risk score and target management strategy.

[0016] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described knowledge graph-based credit risk customer management method.

[0017] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the aforementioned knowledge graph-based credit risk customer management method.

[0018] In the aforementioned solution based on knowledge graphs for credit risk customer management, the basic customer information and credit information of the customer to be managed can be obtained. Risk event analysis can be performed on the basic customer information to obtain first risk event characteristic information, and risk information extraction can be performed on the credit information to obtain second risk event characteristic information. Knowledge fusion processing can then be performed on the first and second risk event characteristic information to construct a credit risk knowledge graph. Based on this knowledge graph, the credit risk of the customer to be managed can be analyzed and evaluated to obtain a more accurate and comprehensive target credit risk score and a more appropriate and customer-aligned target management strategy, thereby effectively improving the accuracy and effectiveness of the credit risk management process. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of an application environment for a knowledge graph-based credit risk customer management method according to an embodiment of the present invention;

[0021] Figure 2This is a flowchart illustrating a knowledge graph-based credit risk customer management method according to an embodiment of the present invention.

[0022] Figure 3 This is a schematic diagram of a knowledge graph-based credit risk customer management device according to an embodiment of the present invention;

[0023] Figure 4 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention;

[0024] Figure 5 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0027] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.

[0028] Currently, existing financial credit risk identification methods can generate a financial credit knowledge graph based on the borrower's information and a financial event graph based on financial event data and financial domain ontology. Further, based on the borrower's information, the financial credit knowledge graph, and the financial event graph, a financial credit risk identification model is generated through machine learning. This model updates the financial event graph based on real-time financial events. Based on the updated financial event graph and the information of the borrower to be predicted, the financial credit risk identification model outputs the financial credit risk information of the real-time financial event for the borrower to be predicted. However, this method focuses more on predicting financial credit risk through the constructed financial credit risk identification model to improve the predictability of credit risk for credit entities. It does not effectively integrate and utilize existing financial data, and therefore cannot achieve precise credit risk management.

[0029] The knowledge graph-based credit risk customer management method provided in this invention can be applied to, for example... Figure 1 In this application environment, the client communicates with the server via a network. For example, taking a scenario where credit risk customer management is performed based on a knowledge graph using relevant information of the customers to be managed, the target user can upload the basic customer information and credit information of the customers to be managed through the client. The target user can be a manager or R&D personnel performing credit risk customer management tasks; this application does not impose any restrictions on this. Correspondingly, the server can obtain the basic customer information and credit information of the customers to be managed through the client, perform risk event analysis on the basic customer information to obtain first risk event feature information, and extract risk information from the credit information to obtain second risk event feature information. The first and second risk event feature information can then be fused to construct a credit risk knowledge graph. Based on this knowledge graph, the credit risk of the customers to be managed can be analyzed and evaluated to obtain a target credit risk score and target management strategy, which are then fed back to the client. The client can receive the target credit risk score and target management strategy from the server and display them on the client for the target user to query or browse. By adopting the knowledge graph-based credit risk customer management method provided in this application, existing credit data can be effectively integrated and utilized to improve the accuracy of the credit risk management process.

[0030] The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will now be described in detail through specific embodiments.

[0031] Please see Figure 2 As shown, Figure 2 A flowchart illustrating a knowledge graph-based credit risk customer management method provided in an embodiment of the present invention includes the following steps:

[0032] S10 obtains basic customer information and customer credit information of customers to be managed.

[0033] Among these, "customers to be managed" can refer to customers awaiting credit risk management, i.e., customers for whom credit risk scoring has not yet been conducted and for whom credit risk management strategies have not yet been developed. Customer basic information can include, but is not limited to, the customer's name, gender, age, marital status, education level, contact information, residential address, occupation, guarantor information, and litigation information.

[0034] The client's occupational information may include, but is not limited to, the client's job title, salary and benefits, working hours, company name, company type, company size, industry, company address, and legal representative, etc. This application does not impose any restrictions on this.

[0035] Information related to the client guarantor may include, but is not limited to, the client guarantor's name, gender, contact information, address, occupation, assets, credit history, and relationship between the client and the client. This application does not impose any restrictions on this.

[0036] Information related to client litigation may include, but is not limited to, the number of client litigation cases, the types of client litigation cases, the details of client litigation cases, the client's litigation status, the duration of client litigation, the amount involved in client litigation cases, the judgment results of client litigation cases, and related legal documents of client litigation cases. This application does not impose any restrictions on this.

[0037] Customer credit information can be credit-related information generated after a customer under management has engaged in credit activities. This customer credit information may include, but is not limited to, customer credit application records, customer repayment records, customer credit limits, customer loan balances, customer loan purposes, and customer credit institutions, etc., and this application does not impose any restrictions on this.

[0038] S20: Analyze the customer's basic information for risk events to obtain the characteristic information of the first risk event.

[0039] The first risk event characteristic information can be the risk event characteristic information obtained after performing risk event analysis on the customer's basic information. Risk event analysis can be based on the customer's basic information to explore and evaluate risk events from multiple dimensions. For example, the server can use the customer's occupation-related information in the customer's basic information to explore and evaluate occupational risk events from the perspective of the customer's occupation, and this application does not impose any restrictions on this.

[0040] It is important to understand that performing risk event analysis on basic customer information to obtain the first risk event characteristic information refers to the process of conducting risk event analysis based on basic customer information. Specifically, step S20, which involves performing risk event analysis on basic customer information to obtain the first risk event characteristic information, includes the following steps:

[0041] S21: Embed the customer's basic information to obtain the basic information embedding vector;

[0042] S22: Perform bidirectional risk feature encoding on the basic information embedding vector to obtain the risk feature vector;

[0043] S23: Construct a risk event matrix based on risk feature vectors;

[0044] S24: Perform attention calculation on the risk feature vector and the risk event matrix to obtain the risk attention feature vector;

[0045] S25: Normalize the risk attention feature vector to obtain the first risk event feature information.

[0046] The basic information embedding vector can be the embedding vector obtained after embedding customer basic information. By embedding customer basic information, the server can transform it into data that is easier to process and analyze, while also preserving important features and relationships between data points. This application does not impose any limitations on this. Optionally, the server can use existing embedding methods, such as embedding layers, to obtain the basic information embedding vector; this application does not impose any limitations on this either.

[0047] A risk feature vector can be obtained by bidirectional risk feature encoding of a basic information embedding vector. After embedding the customer's basic information, the server can further extract risk event features from the more easily analyzed and processed basic information embedding vector to obtain a risk feature vector that reflects the likelihood of risk. The server can use bidirectional risk feature encoding to simultaneously consider both positive and negative information in the input data, thereby better capturing the contextual information of the input data and obtaining a more accurate feature vector.

[0048] The server can perform forward feature encoding on the basic information embedding vector to obtain a forward feature vector, and perform reverse feature encoding on the basic information embedding vector to obtain a reverse feature vector. The forward and reverse feature vectors can then be concatenated to obtain a risk feature vector. Optionally, the server can employ a bidirectional encoder, such as a bidirectional recurrent neural network (BRNN), to implement the step of obtaining the risk feature vector; this application does not impose any limitations on this approach.

[0049] Optionally, the server performs forward feature encoding on the basic information embedding vector to obtain a forward feature vector. The process can be seen in the following formula:

[0050] h P1 =f1(w P11 h P1-1 +w P12 c i +b P1 )

[0051] Among them, h P1 f1 can be used to represent a positive feature vector; f1 can be used to represent a positive activation function; w P11 This can be used to represent the first positive weight matrix, which can be the weight matrix corresponding to the eigenvector of the previous positive position; h P1-1 It can be used to represent the feature vector of the previous positive position; w P12 This can be used to represent the second positive weight matrix, which can be the weight matrix corresponding to the current input data (i.e., the basic information embedding vector); c i It can be used to represent basic information embedding vectors; b P1 It can be used to represent a positive bias.

[0052] Optionally, the server performs inverse feature encoding on the basic information embedding vector to obtain the inverse feature vector. The process can be seen in the following formula:

[0053] h P2 =f2(wP21 h P2-1 +w P22 c i +b P2 )

[0054] Among them, h P2 f2 can be used to represent inverse feature vectors; f2 can be used to represent inverse activation functions; w P21 This can be used to represent the first inverse weight matrix, which can be the weight matrix corresponding to the eigenvector at the previous inverse position; h P2-1 It can be used to represent the feature vector of the previous reverse position; w P22 This can be used to represent the second inverse weight matrix, which can be the weight matrix corresponding to the current input data (i.e., the basic information embedding vector); c i It can be used to represent basic information embedding vectors; b P2 It can be used to represent reverse bias.

[0055] Optionally, the server-side concatenates the positive and negative feature vectors to obtain the risk feature vector, as shown in the following formula:

[0056] c P =[h P1 ,h P2 ]

[0057] Among them, c P It can be used to represent risk feature vectors; h P1 It can be used to represent positive eigenvectors; h P2 [,] can be used to represent inverse feature vectors; [,] can be used to represent concatenation operations.

[0058] The risk event matrix can be constructed based on risk feature vectors. It's important to understand that constructing the risk event matrix based on risk feature vectors refers to the process of generating the risk event matrix. Specifically, step S23, which involves constructing the risk event matrix based on risk feature vectors, includes the following steps:

[0059] S231: Obtain the first risk event trigger vector and the first context vector corresponding to the first risk event trigger vector from the risk feature vector;

[0060] S232: Obtain the first risk event argument vector and the second context vector corresponding to the first risk event trigger vector from the risk feature vector;

[0061] S233: Concatenate the first risk event trigger vector, the first context vector, the first risk event argument vector, and the second context vector to obtain the first risk event vector;

[0062] S234: Obtain the second risk event trigger vector and the third context vector corresponding to the second risk event trigger vector from the risk feature vector;

[0063] S235: Obtain the second risk event argument vector corresponding to the second risk event trigger vector and the fourth context vector corresponding to the second risk event argument vector from the risk feature vector;

[0064] S236: Concatenate the second risk event trigger vector and the third context vector, as well as the second risk event argument vector and the fourth context vector, to obtain the second risk event vector;

[0065] S237: Construct a risk event matrix based on the first risk event vector and the second risk event vector.

[0066] The first risk event trigger vector can be the vector corresponding to the first risk trigger word in the risk feature vector. Risk event trigger words can include, but are not limited to, abnormal guarantor, unstable income, industry risks, litigation, and suspected fraud, etc., and this application does not impose any restrictions on them.

[0067] The first context vector can be the vector corresponding to the context connected to the first risk event trigger vector. For example, with c t1 Let c be the first risk event trigger vector in the risk feature vector. Then, the preceding one or more risk feature vectors connected to this first risk event trigger vector (e.g., referred to as c) t1-1 ), and one or more subsequent risk feature vectors (such as c) connected to the first risk event trigger vector. t1+1 ), can be the aforementioned first context vector, and this application does not impose any restrictions on it.

[0068] The first risk event argument vector can be the vector corresponding to the various entities or elements associated with the first risk trigger word that participate in constituting the event. The second context vector can be the vector corresponding to the context connected to the first risk event argument vector. For example, with c a1 Let be the first risk event argument vector in the risk feature vector. Then, the preceding one or more risk feature vectors (such as c) connected to this first risk event argument vector... a1-1 ), and one or more subsequent risk feature vectors (such as c) connected to the first risk event argument vector. a1+1 ), can be the aforementioned second context vector, and this application does not impose any restrictions on it.

[0069] The first risk event vector can be obtained by concatenating the first risk event trigger vector, the first context vector, the first risk event argument vector, and the second context vector.

[0070] The second risk event trigger vector can be the vector corresponding to the second risk trigger word in the risk feature vector. The third context vector can be the vector corresponding to the context connected to the second risk event trigger vector. The second risk event argument vector can be the vector corresponding to the various entities or elements associated with the second risk trigger word that participate in constituting the event. The fourth context vector can be the vector corresponding to the context connected to the second risk event argument vector.

[0071] The second risk event vector can be obtained by concatenating the second risk event trigger vector, the third context vector, the second risk event argument vector, and the fourth context vector. Optionally, the server can also obtain other risk trigger words (such as the third risk trigger word, the fourth risk trigger word, etc.) from the risk feature vector to further determine the risk event vectors (such as the third risk event vector, the fourth risk event vector, etc.) corresponding to other risk trigger words. This application does not impose any restrictions on this.

[0072] The server can construct a risk event matrix based on the first and second risk event vectors. Optionally, the process of the server constructing the risk event matrix based on the first and second risk event vectors can be seen in the following formula:

[0073]

[0074] Among them, E s It can be used to represent a risk event matrix; e s1 It can be used to represent the first risk event vector; e s2 It can be used to represent the second risk event vector; c t1-1 It can be used to represent the preceding risk feature vector connected to the first risk event argument vector, such as the first context vector in the first context vector; c t1 It can be used to represent the first risk event trigger vector; c t1+1 It can be used to represent a subsequent risk feature vector connected to the first risk event argument vector, such as the first context vector in the first context vector; c a1-1 It can be used to represent the preceding risk feature vector connected to the first risk event argument vector, such as the second context vector in the second context vector; c a1 It can be used to represent the argument vector of the first risk event; c a1+1It can be used to represent a subsequent risk feature vector connected to the first risk event argument vector, such as the second context vector in the second context vector; c t2-1 It can be used to represent the preceding risk feature vector connected to the second risk event argument vector, such as the third context vector in the third context vector; c t2 It can be used to represent the second risk event trigger vector; c t2+1 It can be used to represent a subsequent risk feature vector connected to the second risk event argument vector, such as the third context vector in the third context vector; c a2-1 It can be used to represent the preceding risk feature vector connected to the second risk event argument vector, such as the fourth context vector in the fourth context vector; c a2 It can be used to represent the argument vector of the second risk event; c a2+1 It can be used to represent a subsequent risk feature vector connected to the second risk event argument vector, such as the fourth context vector in the fourth context vector.

[0075] For steps S231-S237, the server obtains the first risk event trigger vector and the corresponding first context vector from the risk feature vector. It can further obtain the first risk event argument vector and the corresponding second context vector from the risk feature vector. The server then concatenates the first risk event trigger vector and the first context vector, as well as the first risk event argument vector and the second context vector, to obtain the first risk event vector. Finally, by obtaining the second risk event trigger vector and the corresponding third context vector from the risk feature vector, it can further... The first step involves obtaining the second risk event argument vector and the fourth context vector corresponding to the second risk event trigger vector from the risk feature vector. Then, the second risk event trigger vector, the third context vector, the second risk event argument vector, and the fourth context vector are concatenated to obtain the second risk event vector. Based on the first and second risk event vectors, a risk event matrix can be constructed. This facilitates the determination of more accurate first risk event feature information in subsequent steps, effectively improving the accuracy of risk scoring for clients under management and enhancing the accuracy of the credit risk management process.

[0076] It is important to understand that performing attention calculation on the risk feature vector and risk event matrix to obtain the first risk event feature information refers to the process of performing attention calculation using an attention mechanism. Specifically, step S24, which involves performing attention calculation on the risk feature vector and risk event matrix to obtain the first risk event feature information, includes the following steps:

[0077] S241: Calculate the first risk attention score based on the first risk event trigger vector and the risk event matrix in the risk feature vector;

[0078] S242: Perform a weighted summation operation between the first risk attention score and the risk trigger vector in the risk feature vector to obtain the first risk attention feature vector;

[0079] S243: Calculate the second risk attention score based on the second risk event trigger vector and the risk event matrix in the risk feature vector;

[0080] S244: Perform a weighted summation operation between the second risk attention score and the risk trigger vector in the risk feature vector to obtain the second risk attention feature vector;

[0081] S245: Normalize the first risk attention feature vector and the second risk attention feature vector to obtain the first risk event feature information.

[0082] The first risk attention score can be the attention score obtained by calculating the attention score of the first risk event trigger vector and the risk event matrix in the risk feature vector. The first risk attention feature vector can be the attention feature vector obtained by performing a weighted summation operation on a risk attention score and other risk trigger vectors in the risk feature vector.

[0083] Optionally, the server can calculate the first risk attention score based on the first risk event trigger vector and the risk event matrix in the risk feature vector. The process of obtaining the first risk attention score can be seen in the following formula:

[0084] atten score1 =softmax(c t1 T w1 T w2E s )

[0085] Among them, atten score1 It can be used to represent the first risk attention score; softmax can be used to represent the normalization operation performed by the softmax function; c t1 The vector can be used to represent the first risk event trigger vector, and T can be used to represent the transpose operation; w1 can be used to represent the weight corresponding to the first risk event trigger vector, and T can be used to represent the transpose operation; w2 can represent the weight corresponding to the risk event matrix; E s It can be used to represent a risk event matrix.

[0086] Optionally, the server performs a weighted summation of the first risk attention score with other risk trigger vectors in the risk feature vector to obtain the first risk attention feature vector. See the following formula for the process:

[0087]

[0088] Where F1 can be used to represent the first risk attention feature vector; i can be used to represent the index of the risk feature vector; n can be used to represent the number of risk feature vectors; δ1 can be used to represent the attention weight corresponding to the first risk attention score; atten score1 It can be used to represent the first risk attention score; c ti It can be used to represent the risk trigger vector in the risk feature vector.

[0089] The second risk attention score can be obtained by calculating the attention score of the second risk event trigger vector and the risk event matrix in the risk feature vector. The second risk attention feature vector can be obtained by performing a weighted summation operation between the second risk attention score and the risk trigger vector in the risk feature vector.

[0090] Optionally, the server can calculate the second risk attention score based on the second risk event trigger vector and the risk event matrix in the risk feature vector. The process of obtaining the second risk attention score can be seen in the following formula:

[0091] atten score2 =softmax(c t2 T w3 T w4E s )

[0092] Among them, atten score2 It can be used to represent the second risk attention score; softmax can be used to represent the normalization operation performed by the softmax function; c t2 The vector can be used to represent the second risk event trigger vector, and T can be used to represent the transpose operation; w3 can be used to represent the weights corresponding to the second risk event trigger vector, and T can be used to represent the transpose operation; w4 can be used to represent the weights corresponding to the risk event matrix; E s It can be used to represent a risk event matrix.

[0093] Optionally, the server can perform a weighted summation of the second risk attention score and the risk trigger vector in the risk feature vector to obtain the second risk attention feature vector. See the following formula for the process:

[0094]

[0095] Where F2 can be used to represent the second risk attention feature vector; i can be used to represent the index of the risk feature vector; n can be used to represent the number of risk feature vectors; δ2 can be used to represent the attention weight corresponding to the second risk attention score; atten score2 It can be used to represent the second risk attention score; c ti It can be used to represent the risk trigger vector in the risk feature vector.

[0096] After obtaining the first risk attention feature vector and the second risk attention feature vector, the server can further normalize these vectors to obtain the first risk event feature information. Optionally, the server can also determine risk attention feature vectors (such as the third risk event trigger vector, the fourth risk trigger word, etc.) corresponding to other risk trigger words based on other risk event trigger vectors (such as the third risk attention feature vector, the fourth risk attention feature vector, etc.), thereby determining the first risk event feature information based on each risk attention feature vector. This application does not impose any limitations on this.

[0097] For steps S241-S245, the server calculates an attention score based on the first risk event trigger vector and the risk event matrix in the risk feature vector to obtain a first risk attention score. This first risk attention score is then weighted and summed with the risk trigger vector in the risk feature vector to obtain a first risk attention feature vector. Similarly, the server calculates an attention score based on the second risk event trigger vector and the risk event matrix in the risk feature vector to obtain a second risk attention score. This second risk attention score is then weighted and summed with the risk trigger vector in the risk feature vector to obtain a second risk attention feature vector. This process normalizes the first and second risk attention feature vectors to obtain first risk event feature information. This first risk event feature information is then used in subsequent steps to determine a more accurate target credit risk score, effectively improving the accuracy and effectiveness of the credit risk management process.

[0098] S30: Extract risk information from customer credit information to obtain the second risk event characteristic information.

[0099] The second risk event characteristic information can be the risk event characteristic information obtained after extracting risk information from customer credit information. Risk information extraction can accurately and efficiently identify credit risks based on customer credit information and extract key information related to credit risks. For example, the server can extract relevant key information for further risk analysis based on credit delinquency records, bad credit records, number of loans, lending institution information, and debt limit information in customer credit information. This application does not impose any restrictions on this.

[0100] It is important to understand that determining the server cross-feature set based on the server-related feature set refers to the process of performing feature analysis on each server-related feature in the server-related feature set. Specifically, step S30, which involves determining the server cross-feature set based on the server-related feature set, includes the following steps:

[0101] S31: Preprocess customer credit information to obtain structured credit information;

[0102] S32: Standardize the structured credit information to obtain standard credit information;

[0103] S33: Perform linear correlation calculations based on standard credit information to obtain linear correlation credit information;

[0104] S34: Based on the linearly correlated credit information, determine the set of credit feature values ​​and the set of credit feature vectors;

[0105] S35: Calculate the contribution rate for each credit feature value in the credit feature value set to obtain the current cumulative credit contribution rate;

[0106] S36: Obtain the m target credit feature values ​​corresponding to the cumulative credit contribution rate when it is greater than the preset cumulative contribution rate;

[0107] S37: Determine the feature information of the second risk event based on the m target credit feature vectors corresponding to the m target credit feature values.

[0108] The structured credit information can be obtained by preprocessing customer credit information. This structured credit information can be data tables in a database, data in spreadsheet software, data in a comma-separated values ​​(CSV) file, or data in an Extensible Markup Language (XML) file, etc., and this application does not impose any restrictions on this. Optionally, the structured credit information can be presented in matrix form, and this application does not impose any restrictions on this either.

[0109] This application uses the example of a server-side table of data built based on credit risk keywords for illustration. Optionally, it uses the example of the server extracting risk information from customer credit information based on credit risk keywords such as "overdue, bad record, number of loans, lending institution, and debt amount" for illustration. The server can further transform the extracted information into structured data, and then obtain the second risk event feature information based on the structured data. This application does not impose any restrictions on this.

[0110] For example, the server can convert the extracted information into the following table:

[0111]

[0112] It should be noted that this application uses the credit risk keywords "overdue payments, bad records, number of loans, lending institutions, and debt limits" as examples for illustration, and does not constitute a limitation on this application. Optionally, the customer data in the above table is for illustrative purposes only and does not constitute a limitation on this application.

[0113] Standard credit information can be obtained by standardizing structured credit information. Optionally, taking a structured credit information matrix (such as a structured credit matrix) as an example, the server can calculate the credit mean and standard deviation corresponding to the structured credit matrix. Then, each element in the structured credit matrix is ​​first subtracted from the credit mean of its column, and then divided by the credit standard deviation of its column to obtain standard credit information (such as a standard credit matrix). This eliminates the influence of the mean and makes the data in different columns have the same scale. This application does not impose any restrictions on this.

[0114] Linearly correlated credit information can be the credit information obtained by performing linear correlation calculations on standard credit information. Taking the standard credit information as a standard credit matrix as an example, this linear correlation calculation can be understood as calculating the covariance matrix of the standard credit matrix, and quantitatively evaluating the relationship between the variables represented by different rows (or columns) in the standard credit matrix, thereby obtaining linearly correlated credit information (such as being called a linearly correlated credit matrix). This application does not impose any limitations on this.

[0115] The set of credit feature values ​​may include one or more credit feature values, and the set of credit feature vectors may include one or more credit feature vectors. These credit feature values ​​and feature vectors can be calculated based on linearly correlated credit information. Optionally, the server can determine the set of credit feature values ​​and the set of credit feature vectors based on linearly correlated credit information by calculating the eigenvalue (eig) function of the feature values ​​and feature vectors; this application does not impose any limitations on this.

[0116] The cumulative credit contribution rate can be calculated based on the credit feature values. Optionally, the server can calculate the current cumulative credit contribution rate based on each credit feature value in the set of credit feature values, as shown in the following formula:

[0117]

[0118] in, It can be used to represent the current cumulative credit contribution rate, that is, the cumulative credit contribution rate up to the m-th credit feature value; k can be used to represent the index in the set of credit feature values, and m can be used to represent the number of feature values ​​accumulated up to the current time; γ m It can be used to represent the m-th credit feature value; q can be used to represent the sum of all credit feature values ​​in the set of credit feature values, and q can be used to represent the quantity of all credit feature values ​​in the set of credit feature values.

[0119] The target credit feature value can be one or more credit feature values ​​accumulated when the current cumulative credit contribution rate is greater than the preset cumulative contribution rate. The preset cumulative contribution rate can be a cumulative contribution rate pre-set according to needs. This preset cumulative contribution rate can be pre-set by the target personnel or by the system default setting; this application does not impose any restrictions on this. The target credit feature vector can be a credit feature vector corresponding to the target credit feature value.

[0120] For example, based on the current cumulative credit contribution rate Taking the example of a cumulative contribution rate greater than the preset one, the server can obtain the currently accumulated m credit feature values ​​as the m target credit feature values, and further use the target credit feature vectors corresponding to the m target credit feature values ​​as the m target credit feature vectors, thereby obtaining the second risk event feature information. This application does not impose any restrictions on this.

[0121] For steps S31-S37, the server preprocesses customer credit information to obtain structured credit information, and then standardizes this structured credit information to obtain standard credit information. Linear correlation calculations are then performed based on this standard credit information to obtain linearly correlated credit information. Further, based on this linearly correlated credit information, a set of credit feature values ​​and a set of credit feature vectors are determined. Then, the contribution rate of each credit feature value in the set of credit feature values ​​is calculated to obtain the current cumulative credit contribution rate. This allows for the identification of m target credit feature values ​​corresponding to a preset cumulative contribution rate. Finally, based on the m target credit feature vectors corresponding to these m target credit feature values, more accurate second risk event feature information is determined, effectively improving the accuracy and effectiveness of the credit risk management process.

[0122] S40: Perform knowledge fusion processing on the characteristic information of the first risk event and the characteristic information of the second risk event to construct a credit risk knowledge graph.

[0123] The credit risk knowledge graph can be constructed based on the fused first risk event feature information and the second risk event feature information. For example, the server can use technologies such as entity alignment and relationship matching to merge and integrate similar or identical entities and relationships, and then further fill the fused entities and relationships into a suitable knowledge graph framework, continuously improving and supplementing the relevant attribute information of each entity and relationship, thereby obtaining the credit risk knowledge graph.

[0124] The server can employ suitable knowledge graph construction tools and technologies, such as knowledge graph construction platforms, to create a credit risk knowledge graph framework. It can then connect and organize the entities and relationships within this framework to obtain an accurate and comprehensive credit risk knowledge graph; this application does not impose any restrictions on this. The server can also add detailed annotations and comments to entities and relationships to provide richer information; this application does not impose any restrictions on this. Furthermore, the server can validate the constructed credit risk knowledge graph to check its accuracy and completeness, and optimize and adjust it based on the validation results to improve the quality of the constructed knowledge graph; this application does not impose any restrictions on this.

[0125] Optionally, to simplify calculations and accelerate processing, the server can employ a simple fusion method, such as feature concatenation or weighted averaging, to obtain the fused first risk event feature information and second risk event feature information. Optionally, the server can also employ a more complex fusion method, such as neural network-based fusion or decision tree-based fusion, to more comprehensively and meticulously perform knowledge fusion processing on the first risk event feature information and second risk event feature information; this application does not impose any restrictions on this approach.

[0126] It is understandable that, during the process of integrating and processing knowledge (i.e., the characteristic information of the first risk event and the characteristic information of the second risk event), the server can further reason and expand the knowledge to obtain new knowledge, which can then be displayed in the credit risk knowledge graph. This application does not impose any restrictions on this.

[0127] S50: Based on the credit risk knowledge graph, analyze and evaluate the credit risk of the clients to be managed in order to obtain the target credit risk score and target management strategy.

[0128] The target credit risk score can be a credit risk score obtained after conducting credit risk analysis and evaluation of the client to be managed based on a credit risk knowledge graph. Optionally, the client A to be managed can have different aspects of credit risk scores, such as overdue risk score, high debt risk score, frequent borrowing risk score, etc., which are not restricted in this application.

[0129] The target management strategy can be a management strategy corresponding to the target credit risk score. For example, if there are credit risk scores A, B, and C, then score A can correspond to management strategy A, score B can correspond to management strategy B, and score C can correspond to management strategy C. This application does not impose any restrictions on this. Optionally, the target management strategy can also include different aspects of management strategies, such as management strategies designed for delinquency risk or management strategies designed for frequent borrowing. This application does not impose any restrictions on this.

[0130] As can be seen, in the above scheme, the server can obtain the basic customer information and credit information of the customers to be managed, perform risk event analysis on the basic customer information to obtain the first risk event feature information, and extract risk information from the credit information to obtain the second risk event feature information. Then, the first and second risk event feature information can be fused to construct a credit risk knowledge graph. Based on the credit risk knowledge graph, the credit risk of the customers to be managed can be analyzed and evaluated to obtain a more accurate and comprehensive target credit risk score and a more appropriate target management strategy that is more in line with customer risk management. This can effectively integrate and utilize existing credit data to improve the accuracy of the credit risk management process.

[0131] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. 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 embodiments of the present invention.

[0132] In one embodiment, a knowledge graph-based credit risk customer management device is provided, which corresponds one-to-one with the knowledge graph-based credit risk customer management method described in the above embodiments. For example... Figure 3 As shown, the knowledge graph-based credit risk customer management device includes an acquisition module 101, a first processing module 102, a second processing module 103, a third processing module 104, and a fourth processing module 105. Detailed descriptions of each functional module are as follows:

[0133] Module 101 is used to acquire basic customer information and credit information of customers to be managed.

[0134] The first processing module 102 is used to perform risk event analysis on the customer's basic information to obtain the first risk event characteristic information;

[0135] The second processing module 103 is used to extract risk information from customer credit information to obtain second risk event feature information.

[0136] The third processing module 104 is used to perform knowledge fusion processing on the feature information of the first risk event and the feature information of the second risk event in order to construct a credit risk knowledge graph.

[0137] The fourth processing module 105 is used to analyze and evaluate the credit risk of the client to be managed based on the credit risk knowledge graph, so as to obtain the target credit risk score and target management strategy.

[0138] In one embodiment, the first processing module 102 is specifically used for: embedding customer basic information to obtain a basic information embedding vector; performing bidirectional risk feature encoding on the basic information embedding vector to obtain a risk feature vector; constructing a risk event matrix based on the risk feature vector; and performing attention calculation on the risk feature vector and the risk event matrix to obtain first risk event feature information.

[0139] In one embodiment, the first processing module 102 is specifically configured to: obtain a first risk event trigger vector and a first context vector corresponding to the first risk event trigger vector from the risk feature vector; obtain a first risk event argument vector and a second context vector corresponding to the first risk event trigger vector from the risk feature vector; concatenate the first risk event trigger vector and the first context vector, as well as the first risk event argument vector and the second context vector, to obtain a first risk event vector; obtain a second risk event trigger vector and a third context vector corresponding to the second risk event trigger vector from the risk feature vector; obtain a second risk event argument vector and a fourth context vector corresponding to the second risk event trigger vector from the risk feature vector; concatenate the second risk event trigger vector and the third context vector, as well as the second risk event argument vector and the fourth context vector, to obtain a second risk event vector; and construct a risk event matrix based on the first risk event vector and the second risk event vector.

[0140] In one embodiment, the first processing module 102 is specifically configured to: calculate an attention score based on a first risk event trigger vector and a risk event matrix in the risk feature vector to obtain a first risk attention score; perform a weighted summation operation on the first risk attention score and the risk trigger vector in the risk feature vector to obtain a first risk attention feature vector; calculate an attention score based on a second risk event trigger vector and a risk event matrix in the risk feature vector to obtain a second risk attention score; perform a weighted summation operation on the second risk attention score and the risk trigger vector in the risk feature vector to obtain a second risk attention feature vector; and normalize the first risk attention feature vector and the second risk attention feature vector to obtain first risk event feature information.

[0141] In one embodiment, the second processing module 103 is specifically used for: preprocessing customer credit information to obtain structured credit information; standardizing the structured credit information to obtain standard credit information; performing linear correlation calculation based on the standard credit information to obtain linearly correlated credit information; determining a set of credit feature values ​​and a set of credit feature vectors based on the linearly correlated credit information; calculating the contribution rate of each credit feature value in the set of credit feature values ​​to obtain the current cumulative credit contribution rate; obtaining m target credit feature values ​​corresponding to when the cumulative credit contribution rate is greater than a preset cumulative contribution rate; and determining second risk event feature information based on the m target credit feature vectors corresponding to the m target credit feature values.

[0142] This invention provides a credit risk customer management device based on a knowledge graph. By acquiring the basic customer information and credit information of the customer to be managed, risk event analysis can be performed on the basic customer information to obtain first risk event feature information, and risk information extraction can be performed on the credit information to obtain second risk event feature information. Then, the first and second risk event feature information can be fused to construct a credit risk knowledge graph. Based on the credit risk knowledge graph, the credit risk of the customer to be managed can be analyzed and evaluated to obtain a more accurate and comprehensive target credit risk score and a more appropriate target management strategy that is more in line with customer risk management. It can effectively integrate and utilize existing credit data to improve the accuracy of the credit risk management process.

[0143] Specific limitations regarding the knowledge graph-based credit risk customer management device can be found in the limitations of the knowledge graph-based credit risk customer management method described above, and will not be repeated here. Each module in the aforementioned knowledge graph-based credit risk customer management device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0144] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the server-side functions or steps of a knowledge graph-based credit risk customer management method.

[0145] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements client-side functions or steps of a knowledge graph-based credit risk customer management method.

[0146] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0147] Obtain basic customer information and customer credit information of customers to be managed;

[0148] Risk event analysis is performed on basic customer information to obtain the characteristic information of the first risk event;

[0149] Risk information is extracted from customer credit information to obtain the characteristic information of the second risk event;

[0150] Knowledge fusion processing is performed on the characteristic information of the first risk event and the characteristic information of the second risk event to construct a credit risk knowledge graph;

[0151] Based on the credit risk knowledge graph, the credit risk of the clients to be managed is analyzed and evaluated to obtain the target credit risk score and target management strategy.

[0152] This invention provides a computer device that, by acquiring basic customer information and credit information of customers to be managed, can perform risk event analysis on the basic customer information to obtain first risk event feature information, and extract risk information from the credit information to obtain second risk event feature information. The first and second risk event feature information can then be fused to construct a credit risk knowledge graph. Based on this knowledge graph, the credit risk of the customers to be managed can be analyzed and evaluated to obtain a more accurate and comprehensive target credit risk score and a more appropriate and customer-aligned target management strategy. This effectively integrates and utilizes existing credit data to improve the accuracy of the credit risk management process.

[0153] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0154] Obtain basic customer information and customer credit information of customers to be managed;

[0155] Risk event analysis is performed on basic customer information to obtain the characteristic information of the first risk event;

[0156] Risk information is extracted from customer credit information to obtain the characteristic information of the second risk event;

[0157] Knowledge fusion processing is performed on the characteristic information of the first risk event and the characteristic information of the second risk event to construct a credit risk knowledge graph;

[0158] Based on the credit risk knowledge graph, the credit risk of the clients to be managed is analyzed and evaluated to obtain the target credit risk score and target management strategy.

[0159] This invention provides a computer-readable storage medium that, by acquiring basic customer information and credit information of a customer to be managed, can perform risk event analysis on the basic customer information to obtain first risk event feature information, and extract risk information from the credit information to obtain second risk event feature information. The first and second risk event feature information can then be fused to construct a credit risk knowledge graph. Based on this knowledge graph, the credit risk of the customer to be managed can be analyzed and evaluated to obtain a more accurate and comprehensive target credit risk score and a more appropriate and effective target management strategy, thereby significantly improving the accuracy and effectiveness of the credit risk management process.

[0160] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0161] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0162] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0163] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A credit risk customer management method based on knowledge graphs, characterized in that, The knowledge graph-based credit risk customer management method includes: Obtain basic customer information and customer credit information of customers to be managed; Risk event analysis is performed on the aforementioned basic customer information to obtain the first risk event characteristic information; Risk information is extracted from the customer credit information to obtain the second risk event feature information; The first risk event feature information and the second risk event feature information are subjected to knowledge fusion processing to construct a credit risk knowledge graph; Based on the credit risk knowledge graph, the credit risk of the client to be managed is analyzed and evaluated to obtain the target credit risk score and target management strategy; The risk event analysis performed on the customer's basic information to obtain first risk event characteristic information includes: The customer's basic information is embedded to obtain a basic information embedding vector; The basic information embedding vector is subjected to forward feature encoding to obtain a forward feature vector; The basic information embedding vector is subjected to inverse feature encoding to obtain an inverse feature vector; The positive feature vector and the negative feature vector are concatenated to obtain the risk feature vector; Construct a risk event matrix based on the aforementioned risk feature vectors; Attention calculation is performed on the risk feature vector and the risk event matrix to obtain the first risk event feature information; The step of extracting risk information from the customer credit information to obtain second risk event feature information includes: The customer credit information is preprocessed to obtain structured credit information; The structured credit information is standardized to obtain standard credit information; Linear correlation calculations are performed based on the standard credit information to obtain linear correlation credit information. Based on the linearly correlated credit information, determine the set of credit feature values ​​and the set of credit feature vectors; The current cumulative credit contribution rate is obtained by calculating the contribution rate of each credit feature value in the set of credit feature values. Obtain the m target credit feature values ​​corresponding to the condition that the cumulative credit contribution rate is greater than the preset cumulative contribution rate; The second risk event feature information is determined based on the m target credit feature vectors corresponding to the m target credit feature values.

2. The knowledge graph-based credit risk customer management method according to claim 1, characterized in that, The construction of the risk event matrix based on the risk feature vector includes: Obtain the first risk event trigger vector and the first context vector corresponding to the first risk event trigger vector from the risk feature vector; Obtain from the risk feature vector a first risk event argument vector corresponding to the first risk event trigger vector and a second context vector corresponding to the first risk event argument vector; The first risk event vector is obtained by concatenating the first risk event trigger vector and the first context vector, as well as the first risk event argument vector and the second context vector. Obtain the second risk event trigger vector and the third context vector corresponding to the second risk event trigger vector from the risk feature vector; Obtain from the risk feature vector the second risk event argument vector corresponding to the second risk event trigger vector and the fourth context vector corresponding to the second risk event argument vector; The second risk event trigger vector and the third context vector, as well as the second risk event argument vector and the fourth context vector, are concatenated to obtain the second risk event vector. Construct a risk event matrix based on the first risk event vector and the second risk event vector.

3. The knowledge graph-based credit risk customer management method according to claim 2, characterized in that, The step of performing attention calculation on the risk feature vector and the risk event matrix to obtain the first risk event feature information includes: The first risk attention score is obtained by calculating the attention score based on the first risk event trigger vector in the risk feature vector and the risk event matrix. The first risk attention score is weighted and summed with the risk trigger vector in the risk feature vector to obtain the first risk attention feature vector. The second risk attention score is obtained by calculating the attention score based on the second risk event trigger vector in the risk feature vector and the risk event matrix. The second risk attention score is weighted and summed with the risk trigger vector in the risk feature vector to obtain the second risk attention feature vector. The first risk attention feature vector and the second risk attention feature vector are normalized to obtain the first risk event feature information.

4. A credit risk customer management device based on knowledge graphs, characterized in that, The knowledge graph-based credit risk customer management device includes: The acquisition module is used to acquire basic customer information and customer credit information of customers to be managed. The first processing module is used to perform risk event analysis on the customer's basic information to obtain first risk event feature information; The second processing module is used to extract risk information from the customer credit information to obtain second risk event feature information. The third processing module is used to perform knowledge fusion processing on the first risk event feature information and the second risk event feature information to construct a credit risk knowledge graph. The fourth processing module is used to analyze and evaluate the credit risk of the customer to be managed based on the credit risk knowledge graph, so as to obtain the target credit risk score and target management strategy. The first processing module is used to perform risk event analysis on the customer's basic information to obtain first risk event feature information, specifically for: The customer's basic information is embedded to obtain a basic information embedding vector; The basic information embedding vector is subjected to forward feature encoding to obtain a forward feature vector; The basic information embedding vector is subjected to inverse feature encoding to obtain an inverse feature vector; The positive feature vector and the negative feature vector are concatenated to obtain the risk feature vector; Construct a risk event matrix based on the aforementioned risk feature vectors; Attention calculation is performed on the risk feature vector and the risk event matrix to obtain the first risk event feature information; The second processing module is used to extract risk information from the customer credit information to obtain second risk event feature information, specifically for: The customer credit information is preprocessed to obtain structured credit information; The structured credit information is standardized to obtain standard credit information; Linear correlation calculations are performed based on the standard credit information to obtain linear correlation credit information. Based on the linearly correlated credit information, determine the set of credit feature values ​​and the set of credit feature vectors; The current cumulative credit contribution rate is obtained by calculating the contribution rate of each credit feature value in the set of credit feature values. Obtain the m target credit feature values ​​corresponding to the condition that the cumulative credit contribution rate is greater than the preset cumulative contribution rate; The second risk event feature information is determined based on the m target credit feature vectors corresponding to the m target credit feature values.

5. The knowledge graph-based credit risk customer management device according to claim 4, characterized in that, The first processing module is used to construct a risk event matrix based on the risk feature vector, specifically for: Obtain the first risk event trigger vector and the first context vector corresponding to the first risk event trigger vector from the risk feature vector; Obtain from the risk feature vector a first risk event argument vector corresponding to the first risk event trigger vector and a second context vector corresponding to the first risk event argument vector; The first risk event vector is obtained by concatenating the first risk event trigger vector and the first context vector, as well as the first risk event argument vector and the second context vector. Obtain the second risk event trigger vector and the third context vector corresponding to the second risk event trigger vector from the risk feature vector; Obtain from the risk feature vector the second risk event argument vector corresponding to the second risk event trigger vector and the fourth context vector corresponding to the second risk event argument vector; The second risk event trigger vector and the third context vector, as well as the second risk event argument vector and the fourth context vector, are concatenated to obtain the second risk event vector. Construct a risk event matrix based on the first risk event vector and the second risk event vector.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the knowledge graph-based credit risk customer management method as described in any one of claims 1 to 3.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the knowledge graph-based credit risk customer management method as described in any one of claims 1 to 3.

Images