Credit management method, system and device based on big data risk control and medium

Abstract

The invention discloses a credit management method, system and device based on big data risk control and a medium, and relates to the technical field of data processing. The method comprises the following steps: acquiring credit information of a plurality of customers; based on the debit and loan data in the credit and loan information, generating a debit and loan relation table among the clients, and based on the guarantee data in the credit and loan information, generating a guarantee relation table among the clients; constructing a customer association degree matrix in combination with the debit and credit relation table and the guarantee relation table among the customers; calculating a risk factor corresponding to each customer according to the customer association degree matrix, and constructing a risk conduction network according to each risk factor; and when the credit behavior of any client has a preset risk event, calculating a risk exposure degree corresponding to a target client associated with the client according to a conduction path of the client in the risk conduction network, and generating early warning information when the risk exposure degree exceeds a preset threshold value. By implementing the technical scheme provided by the invention, the effect of improving the credit management accuracy is achieved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and specifically relates to a credit management method, system, device and medium based on big data risk control. Background Art

[0002] With the rapid development of the financial market and the complexity of lending business, credit management has become increasingly important for the stable operation of banks and financial institutions. Effective credit risk management not only affects the asset quality of financial institutions, but also relates to the security of the entire financial system.

[0003] Currently, existing credit management methods rely on analyzing and warning the credit situation of a single customer to achieve the effect of credit management. However, in actual applications, due to the influence of the correlation between customers on the overall credit risk, it is often difficult to control the overall risk only by performing risk control management on the credit situation of a single customer, thus reducing the accuracy of credit management. Summary of the Invention

[0004] This application provides a credit management method, system, device and medium based on big data risk control, which can improve the accuracy of credit management.

[0005] In a first aspect, this application provides a credit management method based on big data risk control, including: Obtaining credit information of multiple customers; Generating a lending relationship table between each of the customers based on the lending data in each of the credit information, and generating a guarantee relationship table between each of the customers based on the guarantee data in each of the credit information; Combining the lending relationship table and the guarantee relationship table between each of the customers to construct a customer correlation matrix; Calculating a risk factor corresponding to each of the customers according to the customer correlation matrix, and constructing a risk conduction network according to each of the risk factors; When a preset risk event exists in the credit behavior of any customer, calculating a risk exposure corresponding to a target customer associated with the customer according to the conduction path of the customer in the risk conduction network, and generating a warning message when the risk exposure exceeds a preset threshold.

[0006] In a second aspect of this application, there is provided a credit management system based on big data risk control, the system includes: An information acquisition module for acquiring credit information of multiple customers; The correlation matrix determination module is used to generate a lending relationship table among the customers based on the lending data in each of the credit information, and generate a guarantee relationship table among the customers based on the guarantee data in each of the credit information; combine the lending relationship table and the guarantee relationship table among the customers to construct a customer correlation matrix; The risk conduction network determination module is used to calculate the risk factors corresponding to each of the customers according to the customer correlation matrix, and construct a risk conduction network according to each of the risk factors; The risk early warning module is used to, when a preset risk event exists in the credit behavior of any customer, calculate the risk exposure degree corresponding to the target customer associated with the customer according to the conduction path of the customer in the risk conduction network, and generate a warning message when the risk exposure degree exceeds a preset threshold.

[0007] In the third aspect of the present application, an electronic device is provided, including a memory, a processor, and a program stored on the memory and executable on the processor. When the program is loaded and executed by the processor, it can implement a credit management method based on big data risk control.

[0008] In the fourth aspect of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor is enabled to implement a credit management method based on big data risk control.

[0009] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: By adopting the above technical solutions, the credit information of multiple customers is obtained, and a lending relationship table and a guarantee relationship table are respectively generated based on the lending data and the guarantee data, which can comprehensively reflect the business transaction relationships among the customers. By combining the lending relationship table and the guarantee relationship table to construct a customer correlation matrix, the correlation degree among the customers can be quantitatively characterized. Furthermore, based on the customer correlation matrix, the risk factors corresponding to each customer are calculated, and a risk conduction network is constructed, so that the system can accurately depict the conduction characteristics of risks in the customer group. When it is detected that any customer has a preset risk event, the system will calculate the risk exposure degree of the target customer associated with the customer according to the conduction path of the customer in the risk conduction network, and generate a warning message in time when the risk exposure degree exceeds a preset threshold, thereby realizing the early identification and warning of potential risks. This credit management method breaks through the limitation of only focusing on the credit status of a single customer in the traditional way, can effectively predict and prevent the chain risks caused by the correlation among customers, and improves the accuracy rate of credit risk management. Description of the Drawings

[0010] Figure 1It is a schematic flowchart of a credit management method based on big data risk control provided by an embodiment of the present application; Figure 2 It is a schematic structural diagram of a credit management system based on big data risk control provided by an embodiment of the present application; Figure 3 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application.

[0011] Explanation of reference numerals: 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. Detailed implementation manners

[0012] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments.

[0013] In the description of the embodiments of the present application, words such as "for example" or "for illustration" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "for example" or "for illustration" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, the use of words such as "for example" or "for illustration" is intended to present relevant concepts in a specific manner.

[0014] In the description of the embodiments of the present application, the meaning of the term "a plurality" refers to two or more. For example, a plurality of systems refers to two or more systems, and a plurality of screen terminals refers to two or more screen terminals. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly indicating the technical features indicated. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. The terms "include", "comprise", "have" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0015] The embodiments of the present application provide a credit management method based on big data risk control. In one embodiment, please refer to Figure 1 , Figure 1 It is a schematic flowchart of the credit management method based on big data risk control provided by the embodiments of the present application. This method can be implemented depending on a computer program, which can be integrated in an application or run as an independent tool application. This method can also be implemented depending on a single-chip microcomputer and can also run on a credit management system based on the von Neumann architecture and based on big data risk control. Specifically, this method can include the following steps: Step 101: Obtain the credit information of multiple customers.

[0016] Among them, the credit information refers to lending data and guarantee data in the embodiments of this application.

[0017] The lending data includes the loan records of customers, such as information on loan amount, loan term, loan purpose, repayment status, etc. These data reflect the existence of a capital lending relationship among customers.

[0018] The guarantee data includes the contract information of mutual guarantees among customers, such as guarantee amount, guarantee term, types of collateral, etc. These data reflect the risk sharing and transfer behaviors among customers to reduce risks.

[0019] Specifically, obtaining the credit information of multiple customers is the basis of this credit management method based on big data risk control. The credit information of customers includes lending data and guarantee data. Starting from these raw data, a complex association network formed by customers through loans and guarantees can be revealed. The lending data reflects the flow of funds among customers, and the guarantee data reflects the risk sharing situation among customers. Only by comprehensively obtaining the credit information of multiple customers can a real customer relationship map be constructed. Specifically, the lending records and guarantee contract information of customers can be captured from the credit business systems of financial institutions such as banks. These data are usually stored in structured databases, and data capture programs can be written to automatically obtain them. At the same time, it is also necessary to clean, standardize, and integrate data from different sources, eliminate redundancy and noise, and finally form a credit information dataset in a unified format. With the help of big data technology, a large amount of customer credit information can be efficiently stored and processed, laying a foundation for subsequent relationship mining and risk calculation.

[0020] Step 102: Generate a lending relationship table among customers based on the lending data in each credit information, and generate a guarantee relationship table among customers based on the guarantee data in each credit information.

[0021] Among them, the lending relationship table refers to a tabular form that summarizes and records whether there is a direct lending relationship between each pair of customers by analyzing the lending data of multiple customers obtained. It includes the identity information of customer A and customer B, whether there is a lending relationship between customer A and customer B (yes / no), and if so, records the specific lending details such as lending amount and lending term. The lending relationship table reflects the direct capital transactions between different customers as lenders and borrowers.

[0022] The guarantee relationship table refers to a tabular form that summarizes and records whether there is an indirect guarantee relationship between each pair of customers by analyzing the guarantee data of multiple customers obtained. It includes the identity information of customer A and customer B, whether there is a guarantee relationship between customer A and customer B (yes / no), and if so, records the specific guarantee amount, guarantee period, types of guaranteed items, and other guarantee details. The guarantee relationship table reflects the indirect risk correlation situation where different customers provide guarantees to each other to reduce loan risks.

[0023] Specifically, first, construct a lending relationship network diagram based on the lending data. Each customer is regarded as a node, and the lending relationship corresponds to a directed edge, and the edge contains attribute information such as the lending amount. Then, on this basis, introduce new edges in the network according to the guarantee data to represent the guarantee relationship. Finally, a complex network structure model containing two types of relationships is obtained. From this network model, the lending relationship table and the guarantee relationship table can be conveniently extracted. The lending relationship table records whether there is a direct lending relationship between each pair of customers, as well as specific details such as the lending amount and term; the guarantee relationship table records whether there is a guarantee relationship between each pair of customers, the guarantee amount, term, types of guaranteed items, and other information. These two tables are data representations that quantitatively describe the risk transmission paths between customers. By generating these relationship tables, we can accurately calculate the risk correlation intensity between each pair of customers. For example, if there are both lending and guarantee relationships between A and B, then their risk coupling degree will be very high; if there is no relationship between the two, the possibility of risk transmission is relatively small. This refined relationship analysis helps to more accurately assess and warn of risk diffusion, and improve the pertinence and effectiveness of risk control.

[0024] Based on the above embodiments, as an alternative embodiment, in step 102: generating a lending relationship table between each customer based on the lending data in each credit information. This step may further include the following steps: Step 201: Extract the corresponding borrower information, lender information, and transaction information from each lending data; group each customer according to the lender information and borrower information to obtain lending customer groups corresponding to multiple lenders.

[0025] Specifically, first, we extract the identity information of the lender (lender) and the borrower corresponding to each transaction record from the lending data, as well as the transaction details such as the loan amount and term. Then, we use the identity of the lender as the keyword to group all the lending records, and we can get multiple lending customer groups centered on different lenders. Each lending customer group consists of a lender and multiple borrowers who have had lending transactions with it. Within the same lending customer group, there is a potential risk transmission path between different borrowers. Once any of the borrowers defaults, it may affect the lender's fund recovery. The benefit of generating these lending customer groups is that we can more accurately assess the risk exposure faced by each lender. We can analyze the number of borrowers, the overall lending scale, the proportion of single transactions, etc. within a lending customer group, and focus on and rate high-risk areas. At the same time, there is a certain risk correlation between different lending customer groups, because some borrowers may obtain loans from multiple lenders at the same time. This cross-group risk transmission path needs to be reflected in the subsequent correlation matrix and network model. In general, grouping customers according to their identities as lenders and borrowers can refine the calculation granularity of risk exposure, more accurately characterize the risk coupling relationship between different entities, and provide data support for risk management.

[0026] Step 202: Based on each transaction information, calculate the loan intensity between each lender and the corresponding loan customer group; and establish a loan relationship table between each customer according to each loan intensity.

[0027] Specifically, for each generated lending customer group, it is necessary to calculate the lending intensity between the lender and the entire customer group based on all the transaction information contained therein. The lending intensity can be a comprehensive indicator that can take into account the overall lending scale as well as factors such as the lending term and frequency of occurrence. The specific calculation method can be determined according to actual needs. The greater the lending intensity, the higher the risk coupling between the lender and this customer group. If the lending intensity is at a high level, once some borrowers in the customer group default, it may cause a large financial loss impact to the lender. After calculating the lending intensity of all lending customer groups, we can establish a lending relationship table based on this. The rows and columns of this table represent different customer identities, and the values ​​at the cell positions are the lending intensity between the corresponding two customers. Through this lending relationship table, we can not only see whether there is a direct lending relationship between different customers, but also clearly evaluate the strength of risk transmission between them.

[0028] Based on the above embodiment, as an optional embodiment, in step 202: based on each transaction information, the loan strength between each lender and the corresponding loan customer group is calculated. This step may also include the following steps: Step 212: Based on each transaction information, determine the borrowing times and borrowing amounts of each lender and the borrowers in the corresponding borrower group within a preset time period.

[0029] Specifically, for each generated borrower group centered around a lender, we first extract all borrowing transaction information between the lender and each borrower from the transaction records, including borrowing time, amount, etc. Then, after giving a time interval (such as the most recent 1 year), we count the borrowing times and cumulative borrowing amounts between the lender and each borrower within this time period. Through this step, we can not only comprehensively evaluate the overall risk exposure degree between the lender and the entire customer group, but also further refine and analyze the strength of the risk association with each individual borrower. If it is found that a certain borrower has obtained loans from this lender frequently and in large amounts recently, then once this borrower defaults, it will cause a relatively large impact on the lender's capital loss and requires high attention. At the same time, multiple statistics can also be carried out for different time interval lengths, so as to analyze the changing trends of borrowing frequency and scale in the time dimension. This dynamic analysis helps to detect the growth signs of risks in advance and take corresponding measures in a timely manner. After obtaining the borrowing frequencies and amounts of each borrower within each borrower group centered around a lender, we can combine these data with the previously calculated borrowing intensities to establish a more comprehensive and detailed borrowing relationship table. In this table, not only can the overall borrowing intensity among customers be reflected, but also the transaction frequency and amount information included in each specific borrowing relationship can be clearly shown.

[0030] This multi-dimensional and all-round analysis of borrowing relationships can effectively support our accurate identification and quantitative assessment of the risk transmission path. Based on the information in the borrowing relationship table, more precise differential risk response strategies can be formulated, and key attention and control can be implemented for high-risk areas.

[0031] Step 222: Obtain a benchmark borrowing intensity weight combination, which includes weight coefficients corresponding to borrowing times and borrowing amounts.

[0032] Specifically, the benchmark lending intensity weight combination includes two weight coefficients, one corresponding to the weight of the lending frequency and the other corresponding to the weight of the lending amount. The magnitudes of these two weight values reflect the importance ratios they hold when calculating the lending intensity. Usually, this set of weight values will be preset based on historical empirical data or the suggestions of risk experts as the benchmark standard for measuring the lending intensity. For example, the lending frequency weight can be set to 0.4 and the lending amount weight to 0.6, indicating that the impact of the amount is greater. After obtaining the benchmark lending intensity weight combination, we can use it to perform weighted calculations on the lending frequency and lending amount in each obtained lending relationship to get a comprehensive lending intensity. The specific calculation formula is: Lending intensity = Lending frequency weight * Lending frequency + Lending amount weight * Lending amount. In this way, by introducing weights, the risk levels contained in different lending relationships can be quantified more accurately and comprehensively. If only a single factor is considered, it is easy to cause deviations in risk assessment. Of course, the setting of the benchmark weight values can also be adjusted under different risk scenarios. For example, when assessing short-term risks, the lending frequency weight can be appropriately increased; while when assessing long-term risks, more attention should be paid to the impact of the lending amount. This flexibility helps to improve the pertinence and accuracy of risk analysis. By combining multiple dimensions such as lending frequency and amount, and performing weighted summation according to the preset weights, we can finally obtain a set of lending intensity values that can comprehensively reflect the strength of lending relationships. This lays an important data foundation for the subsequent construction of a high-quality lending relationship table and a risk conduction network model.

[0033] Step 232: Based on the benchmark lending intensity weight combination, perform weighted summation on the lending frequency and lending amount corresponding to each lender respectively to obtain the lending intensity between each lender and the corresponding lending customer group.

[0034] Specifically, according to the borrowing times weight and borrowing amount weight in the pre-set benchmark borrowing intensity weight group, multiply the borrowing times by its corresponding weight, multiply the borrowing amount by its corresponding weight, and finally add the two weighted results to obtain a comprehensive borrowing intensity value. Taking a borrowing customer group as an example, assume that the borrowing times weight in the weight group is 0.4 and the borrowing amount weight is 0.6; the borrowing times between this lender and customer A is 5 times, and the borrowing amount is 100,000 yuan; the borrowing times with customer B is 3 times, and the borrowing amount is 150,000 yuan. Then the borrowing intensity with A is 0.4×5 + 0.6×10 = 8; the borrowing intensity with B is 0.4×3 + 0.6×15 = 11. This way of weighted summation can effectively reflect the comprehensive impact of the two dimensions of borrowing times and amount on risk, avoiding the one-sidedness of single-dimensional analysis. By introducing weights, the evaluation results are made more scientific and reasonable. After performing such calculations on all borrowing relationships within each borrowing customer group, the borrowing intensity distribution between this lender and the entire customer group can be obtained. The higher the value, the greater the degree of risk exposure, which requires special attention. These borrowing intensity values can be further combined with the calculated comprehensive borrowing intensity to obtain a complete borrowing relationship table. This table comprehensively depicts the strength of the borrowing relationships between customers, combines the two dimensions of borrowing frequency and amount, and can more accurately evaluate the risk transmission paths between different entities.

[0035] Based on the above embodiments, as an optional embodiment, in step 102: generating a guarantee relationship table between each customer based on the guarantee data in each credit information, this step may further include the following steps: Step 203: Based on each guarantee data, determine the guarantor records and the guaranteed records of each customer; according to each guarantee record and each guaranteed record, determine the guarantee chains between each customer, and the guarantee chain is a sequence of customers with a guarantee and guaranteed relationship.

[0036] Specifically, first, based on the guarantee record data, for each customer, determine their guarantor records and the records of the guaranteed parties respectively. The guarantor records list which other customers this customer has provided guarantees for; while the records of the guaranteed parties list which other customers have provided guarantees for this customer. With these records, we can traverse each customer, starting from it, find the next guaranteed-related customer according to the records of the guaranteed parties, and then continue to find the next one according to the records of the guaranteed parties of that customer, and so on in a loop until no new guaranteed party can be found. In this way, we obtain a complete guarantee chain, representing a sequence of customers with guarantee and guaranteed relationships. Taking customer A as an example, assume that customer B is in the records of the guaranteed parties of A, and customer C is in the records of the guaranteed parties of B, then there is a guarantee chain of A->B->C. Similarly, if there is still customer D in the records of the guaranteed parties of C, then this chain can be extended to A->B->C->D. This kind of guarantee chain actually reflects a risk transmission path. If any link in the chain defaults, then the risk may spread forward or backward along this chain, affecting other relevant customers. Therefore, determining the guarantee chain helps us comprehensively understand the risk correlation among different customers.

[0037] Step 204: Calculate the guarantee amounts between adjacent customers in the guarantee chain; generate a guarantee relationship table based on each guarantee amount.

[0038] Specifically, for each identified guarantee chain, it is necessary to traverse each pair of adjacent customers and find the specific amount of guarantee provided by the guarantor for the guaranteed party in the guarantee data records. For example, for the chain A->B->C, it is necessary to calculate the amount of guarantee provided by A for B and the amount of guarantee provided by B for C respectively. This process can be achieved by writing a query statement and retrieving in the guarantee data table according to the guarantor and the guaranteed party as conditions, so as to obtain the required guarantee amount information. After obtaining the guarantee amounts between all pairs of adjacent guaranteed customers, a guarantee relationship table can be generated. Each row record of this table represents a pair of guarantee relationships, including three fields: the guarantor, the guaranteed party, and the guarantee amount. By establishing the guarantee relationship table, not only can we intuitively see which guarantee relationships exist in the system, but also clearly understand the actual risk exposure amount corresponding to each pair of guarantee relationships, so as to more accurately evaluate its importance. The larger the guarantee amount, the more serious the risk impact that can be generated once a default occurs. With this guarantee relationship table, we can integrate it with the previously constructed lending relationship table to jointly generate a complete risk transmission network model that includes both lending and guarantee risk transmission channels. In this model, the association strength between different nodes can be quantified in an all-round and multi-angle manner, which helps to more accurately evaluate the risk exposure degree of each entity and its potential impact on the entire system.

[0039] Step 103: Combine the loan relationship table and guarantee relationship table between each customer to construct a customer association matrix.

[0040] Among them, the customer correlation matrix refers to a matrix data structure used to quantify and describe the strength of risk correlation between different customers. In the embodiment of the present application, it can be understood as a two-dimensional matrix, in which rows and columns represent different customers, and the values ​​of the matrix elements represent the comprehensive risk correlation between the corresponding two customers. The matrix is ​​used to comprehensively reflect the risk transmission path and its strength between customers within the entire system.

[0041] Specifically, first, based on the loan relationship table, calculate the loan strength value between any two customers. This step has been completed in the previous analysis. It is assumed that the loan strength value calculation method is to use the ratio of the loan amount of one customer to another customer and the total debt of the other customer as the loan strength value between the two. Secondly, based on the guarantee relationship table, calculate the sum of all guarantee amounts between any two customers as guarantors and guarantors. The guarantee relationship table can be screened and summarized through query statements. Then, for each pair of customers, the sum of their loan strength value and guarantee amount is weighted and added to obtain a comprehensive customer association value. The formula is as follows: Customer association value = weight coefficient of loan strength value * loan strength value + weight coefficient of sum of guarantee amount * sum of guarantee amount, where each weight coefficient can be appropriately set by the risk control personnel according to the specific situation. It is assumed that the weight of loan strength value is 0.6 and the weight of sum of guarantee amount is 0.4. Finally, fill the calculated association values ​​between all customer pairs into a matrix, and the rows and columns of the matrix correspond to different customers respectively. The closest risk association exists between Customer B and Customer D, with a correlation value of up to 12; while the correlation between Customer A and D is the smallest, at only 1. This matrix format can intuitively show the risk association network between different customer nodes in the entire system, as well as the overall picture of the strength of the association, which greatly facilitates risk control personnel to quickly identify the risk pain points of the system. For example, if Customer B defaults, it is likely to have a serious risk impact on Customer D; and even if Customer A defaults, the impact on Customer D will be relatively small. With this understanding, we can deploy response plans in advance. At the same time, the matrix can also be used as an important input for building complex risk diffusion models and stress testing models to evaluate the risk resistance of the entire system in situations such as customer defaults.

[0042] Based on the above embodiment, as an optional embodiment, in step 103: combining the loan relationship table and the guarantee relationship table between each customer to construct a customer association matrix, this step may also include the following steps: Step 301: Obtain a relationship weight coefficient group, where the relationship weight coefficient group includes a loan relationship weight coefficient and a guarantee relationship weight coefficient.

[0043] Obtain a set of relationship weight coefficients, including the lending relationship weight coefficient and the guarantee relationship weight coefficient, which are used to weight the impacts of different risk transmission channels when calculating the customer association matrix later.

[0044] The reason for introducing such weight coefficients is that there are differences in the impacts of lending relationships and guarantee relationships on the degree of risk association between customers. Sometimes, the lending relationship is the main channel for leading risk transmission; in other cases, the guarantee relationship may play a greater role. Different weight settings can better reflect the actual situation and improve the accuracy of risk assessment.

[0045] Specifically, risk control personnel or decision-makers can first set an initial set of relationship weight coefficients based on experience. For example: lending relationship weight coefficient = 0.6, guarantee relationship weight coefficient = 0.4. Among them, the weight of the lending relationship is higher because lending behavior is usually the main path of risk transmission. Next, when calculating the customer association matrix, for any two customers, first calculate the lending intensity value and the sum of guarantee amounts between them based on the lending relationship table and the guarantee relationship table respectively. Then, perform weighted summation according to the following formula: customer association value = lending relationship weight coefficient * lending intensity value + guarantee relationship weight coefficient * sum of guarantee amounts. By analogy, the comprehensive association value between any two customers can be obtained and filled into the customer association matrix. This way of weighted summation enables the customer association matrix to comprehensively consider the two different risk transmission paths of lending and guarantee, and comprehensively quantify the risk association intensity between customers. The effect of this step is that by introducing adjustable weight coefficients, we can flexibly adjust the weights of each transmission path according to different risk situations, making the risk assessment results closer to the actual situation, thereby improving the pertinence and effectiveness of decision-making. For example, if the data in a certain period shows that most of the risk events are caused by the guarantee relationship, then the guarantee relationship weight coefficient can be appropriately increased and the lending relationship weight coefficient can be decreased to make the assessment results more in line with the current situation. In addition, in the subsequent model optimization process, we can also iteratively optimize these two weight coefficients based on a large amount of historical data and actual cases to continuously improve the accuracy of risk assessment.

[0046] Step 302: Calculate the lending association degrees between customers based on the lending relationship weight coefficient and the lending intensity in the lending relationship table; calculate the guarantee association degrees between customers based on the guarantee relationship weight coefficient and the guarantee amounts in the guarantee relationship table.

[0047] Specifically, first, based on the lending intensity values between any two customers in the previously generated lending relationship table and the lending relationship weight coefficient obtained in this step, calculate the lending correlation degree between the two, and the formula is as follows: Lending correlation degree = Lending relationship weight coefficient * Lending intensity value. Among them, the lending intensity value can be the ratio of the lent amount to the total debt used previously, or a value obtained by other reasonable calculation methods. Next, based on the sum of all the guarantee amounts between any two customers in the guarantee relationship table as the guarantor and the guaranteed, and the guarantee relationship weight coefficient obtained in this step, calculate the guarantee correlation degree between the two, and the formula is as follows: Guarantee correlation degree = Guarantee relationship weight coefficient * Sum of guarantee amounts. By analogy, the lending correlation degree values and guarantee correlation degree values between any two customers within the system can be obtained.

[0048] The effect of this step is that by quantifying and weighting the two different risk transmission paths of lending and guarantee respectively, it lays a foundation for constructing a comprehensive customer correlation matrix in the subsequent steps. In the next step, we will fuse these two parts of information together to obtain the final customer correlation degree value. This way of splitting and calculating enables us to treat the two risk transmission channels differently and assign corresponding weights according to their different degrees of contribution to the overall risk. This not only avoids the evaluation distortion caused by directly treating them equally, but also can fully consider the influence of both when synthesizing, achieving the comprehensiveness and balance of evaluation. For example, if in some cases, we find that the risk events caused by guarantee behaviors account for a relatively large proportion, then we can appropriately increase the value of the guarantee relationship weight coefficient, thereby increasing the influence weight of the guarantee correlation degree on the overall correlation degree.

[0049] Step 303: Generate corresponding correlation degree vectors according to the lending correlation degrees and guarantee correlation degrees between each customer, and generate a customer correlation matrix according to each correlation degree vector.

[0050] Specifically, first, for any two customers within the system, retrieve the loan correlation degree values and guarantee correlation degree values calculated separately in the previous step. Combine these two values into a two-dimensional vector, which serves as the correlation degree vector between this pair of customers. The first element of the vector is the loan correlation degree, and the second element is the guarantee correlation degree. Secondly, based on the total number of customers N, generate an N*N matrix, where each element of the matrix is a two-dimensional vector. For the element in the i-th row and j-th column of the matrix, assign it the correlation degree vector calculated between customer i and customer j. Finally, set the diagonal elements to zero vectors because the correlation degree between the same customer and itself is meaningless. Through this step, a complete customer correlation degree matrix can be constructed, where each element is a two-dimensional vector, respectively showing the risk correlation intensity between any two customers in the two dimensions of loan and guarantee, as well as their overall correlation degree. The effect of this step is that by integrating information on multi-dimensional risk conduction channels, the customer correlation degree matrix can comprehensively quantify and display the complex risk correlation network among customers within the entire system, providing a clear basis for risk control personnel to identify the pain points of system risks. At the same time, this matrix can also provide important inputs for constructing complex risk conduction models and stress test models, helping to evaluate and predict the overall risk resistance ability of the system.

[0051] Step 104: According to the customer correlation degree matrix, calculate the risk factors corresponding to each customer, and construct a risk conduction network based on each risk factor.

[0052] Among them, the risk factor can be understood as an indicator used to quantify and evaluate the impact degree of each customer node on the entire risk network. The magnitude of the risk factor directly reflects the overall risk correlation intensity between this customer and other customers within the entire system. The larger the risk factor value, the stronger the risk connection between this customer and more other customers, and once a risk event occurs, the higher the likelihood of its triggering the spread of system risks.

[0053] The risk conduction network can be understood as a risk model that represents the customer nodes and their complex risk correlation relationships in a network topology structure. In this network model, the size or color depth of the node corresponds to the risk factor value of this customer, and the larger the value, the greater the impact of this node on the network; the connection weight between any two nodes is equal to the element value of the correlation degree vector corresponding to these two customers in the customer correlation degree matrix, which is used to depict the path and intensity of risk conduction between the two.

[0054] Specifically, first, traverse each row in the customer correlation matrix, sum all the elements of the row vector, and the obtained total value is the risk factor value of the customer. The basis for this approach is that the higher the degree of association between a customer and other customers, the more paths there are for risk initiation and transmission, and the greater the impact on the overall network. Then, sort the risk factor values of all customers from largest to smallest, and identify the customers with larger risk factor values, which correspond to the core risk nodes within the system. Once a risk event occurs, it may trigger a comprehensive risk spread. Next, combine the customer correlation matrix and the risk factor values to construct the entire risk transmission network model. The specific method is to use each customer as a node, and the size or color of the node corresponds to its risk factor value. The connection weight between any two nodes is taken from the corresponding correlation vector element value in the customer correlation matrix. Finally, visually display the constructed risk transmission network to intuitively present the topological structure and key features of the entire network. Through this step, the previously obtained risk association information can be transformed into a risk model with hierarchical and structural features, providing an overall insight into and quantification of the complex risk associations and transmission paths within the system, identifying key risk nodes and weak areas, and laying a foundation for subsequent risk stress testing, scenario simulation, and other analyses.

[0055] Based on the above embodiments, as an alternative embodiment, in step 104: According to the customer correlation matrix, calculate the risk factors corresponding to each customer, and based on each risk factor, construct a risk transmission network. This step may further include the following steps: Step 401: Obtain the historical default records of each customer; based on each historical default record and the customer correlation matrix, determine the initial risk factor of each customer.

[0056] Specifically, first, count the number of defaults and the default amount of each customer in history to construct their default records. This data can usually be obtained from credit rating agencies or the internal credit information system of banks. Secondly, set a default scoring mechanism to calculate the weighted sum of the number of defaults and the amount of each customer to obtain a comprehensive default score value. The higher the score value, the more serious the customer's past default behavior and the higher the potential risk level. Then, referring to the customer correlation matrix, combine the default score value with the association strength between customers to determine the initial risk factor value of each customer. The calculation method can adopt models such as weighted average, so that the risk factor of a customer is not only determined by its own default record but also affected by the default situations of other customers associated with it. Finally, organize all the obtained initial risk factor values of customers to generate a complete risk factor list, which is used as the basic data input for constructing the subsequent risk transmission network model.

[0057] Step 402: Calculate the risk conduction coefficient corresponding to each customer according to the preset risk attenuation coefficient and the initial risk factors of each customer; construct a risk conduction network with each customer as a node and each risk conduction coefficient as the edge weight.

[0058] Specifically, first, according to historical data, a risk attenuation coefficient is preset in advance. This coefficient reflects the natural attenuation rate of risk during conduction in the network, and its value ranges from 0 to 1. The closer it is to 0, the faster the attenuation during risk conduction. Then, traverse each row vector of the customer correlation matrix. Taking the customer node of this row vector as the starting point, multiply its initial risk factor value by the power of the preset risk attenuation coefficient. The result obtained is the risk conduction coefficient corresponding to this customer when conducting risk to other customer nodes. Next, organize all the customer risk conduction coefficients obtained and input them as the basic data for constructing the risk conduction network model. In this network model, customers are used as nodes, and the edge weight between any two nodes is the risk conduction coefficient when one node conducts risk to another node. Finally, visually display the constructed risk conduction network to intuitively present the path topological structure of the internal risk conduction of the entire network. Through this step, the conduction process of risk within the network can be reasonably quantified, and a risk conduction network model that can depict the strength of risk diffusion can be constructed. This enables us to more accurately simulate the impact degree and scope of a single-node risk event on other nodes in the network during stress testing, scenario simulation, etc., and evaluate the risk resistance ability of the overall system. At the same time, the introduction of the risk conduction coefficient also provides a quantitative basis for subsequent network analysis work such as key path identification and risk concentration area discovery. Combining network analysis algorithms, it can be found which risk conduction paths are the most critical and which areas are the weak points where risks are most likely to concentrate, so as to formulate targeted countermeasures. By calculating the risk conduction coefficient and constructing the risk conduction network model, the risk information of a single customer node can be transformed into the dynamic process of internal risk diffusion in the entire network, so as to more accurately quantify and depict the complex risk correlation paths and influence ranges, laying a model foundation for comprehensive risk analysis and control.

[0059] Step 105: When there is a preset risk event in the credit behavior of any customer, calculate the risk exposure corresponding to the target customer associated with the customer according to the conduction path of the customer in the risk conduction network, and generate a warning message when the risk exposure exceeds the preset threshold.

[0060] Among them, the preset risk events can be understood as a series of abnormal customer behaviors or states that trigger risk conduction analysis, preset based on historical experience or regulatory requirements. The preset risk events refer to situations such as default, insolvency, credit rating downgrade, and severe capital liquidity difficulties for customers of financial institutions such as banks, securities companies, and insurance companies, which can all be regarded as the occurrence of risk events. These abnormal conditions often reflect problems in the customer's operation or financial situation, and their own credit risks, default risks, etc. have emerged, and it is very likely to trigger further conduction and diffusion of risks among related customers.

[0061] The conduction path refers to the route that risk sequentially passes through when starting from the source customer where a risk event occurs and through the connected risk conduction edges in the network to other associated customer nodes.

[0062] The risk exposure can be understood as the degree of potential risk impact faced by other connected customers when a risk event occurs to a certain customer, quantified based on its position and association relationship in the risk conduction network. Specifically, the risk exposure refers to the magnitude of the secondary risk impact caused by the source risk event to related customers through the risk conduction path.

[0063] Specifically, first, identify the customer with the occurred risk event, such as a customer's default behavior. Then, find the position and conduction path of this customer in the risk conduction network model. Through the previously calculated risk conduction coefficient, we can calculate the conduction impact of risk in the network level by level along the path direction starting from this customer. For the next-level customer nodes directly connected to this customer, calculate their risk exposure as the default scale of this customer multiplied by the corresponding risk conduction coefficient; for the indirectly connected customers at the next level, their risk exposure is equal to the exposure of the upper-level customer multiplied by the corresponding conduction coefficient, and so on. Traverse all target customers in the network that have a conduction path with the event customer and calculate their corresponding risk exposure values. Next, set a risk exposure threshold. When the risk exposure of any target customer exceeds this threshold, it can be determined that this customer is significantly affected by the risk. At this time, the system will generate corresponding warning information to remind the monitoring personnel to take countermeasures. Finally, record and archive the results of exposure calculation and warning generation, and at the same time maintain and update the risk conduction network model to prepare for the next round of analysis. Through this step, it is possible to automatically and quickly calculate all potentially affected customers and their risk exposure degrees according to the established risk conduction network at the first time when a major risk event occurs to a customer, issue warnings to high-risk customers, and thus minimize the further spread of risks in the network.

[0064] Based on the above embodiments, as an alternative embodiment, in step 105: According to the conduction path of the customer in the risk conduction network, calculate the risk exposure of the target customer associated with the customer. This step may further include the following steps: Step 501: Randomly generate multiple conduction paths in the risk conduction network.

[0065] Specifically, first, determine the purpose and requirements for randomly generating the conduction path, such as for historical case review, specific scenario stress testing, or comprehensive path simulation. Then, in the set of nodes and edges of the entire risk conduction network model, gradually generate qualified conduction paths through random sampling or sampling methods based on probability distributions. These paths may be in the form of loops or non-cyclic paths. During the generation process, different constraint conditions can be set, such as path length, conduction coefficient threshold, customer attribute requirements, etc., to ensure that the generated paths have specific conduction characteristics or risk impacts. Next, specify various entry conditions for each generated conduction path, such as the source customer where the risk event occurs, the default scale, the probability of the event occurring, etc., and different risk diffusion parameters can be set. Then, based on the above settings, simulate the spread process after the risk event occurs in the risk conduction network model, and calculate the impact results such as the risk exposure of each target customer. Finally, summarize and analyze the simulation results, and output detailed reports on the risk conduction path, impact degree, etc., for evaluating the effectiveness of the existing model and strategy, or supporting the formulation of new strategies.

[0066] Through this step, the risk conduction network model can be fully utilized, and by randomly generating conduction paths, multi-dimensional simulations can be carried out for various possible risk scenarios to test the pressure-bearing capacity of the system under extreme conditions. This not only helps to verify the accuracy and applicability of the model, but more importantly, enables us to identify the weak links and potential crises in the system in advance before the risk event actually occurs, providing a basis for timely formulating response plans.

[0067] Step 502: Obtain the occurrence frequency of the target customer in each conduction path; multiply the occurrence frequency of the target customer in each conduction path by the customer's risk factor to obtain the corresponding initial risk exposure.

[0068] Specifically, first, traverse all the randomly generated conduction paths. For each path, count the number of times the target customers appear in it to obtain a frequency value. Then, for each target customer, multiply the frequency values of that customer in all paths by the corresponding risk factor value of the customer. The risk factor here can be indicators reflecting the customer's own risk level, such as the default rate, asset-liability ratio, cash flow status, etc. By multiplying the path frequency by the risk factor, we obtain the initial risk exposure value for each customer. The larger this value is, the higher the risk correlation of the customer in the network, the worse the customer's own risk situation, and the greater the potential risk exposure. Next, a risk exposure threshold can be set, and customers with an initial exposure higher than this threshold are listed as key attention objects, and risk control measures are preferentially implemented. At the same time, the initial exposure values can also be normalized to map the risk exposure levels of all customers to the same magnitude for easy comparison and ranking among each other.

[0069] Through this step, the network model is combined with the customer risk attributes, a preliminary assessment of the conduction effect of systemic risk is made, a group of high-risk customers are identified, and a foundation for more refined analysis in the future is also laid.

[0070] Step 503: Calculate the average value of each initial risk exposure to obtain the risk exposure corresponding to the target customer.

[0071] Specifically, first, for each target customer, we first collect the initial risk exposure values of that customer in all randomly generated conduction paths into a set. Next, for this set of the customer, we can calculate statistical averages such as the arithmetic mean, weighted mean, median, geometric mean, etc. as the final risk exposure assessment result for that customer. In actual operation, the arithmetic mean can be used first to simply average all the initial exposure values. If individual extreme values have a greater impact on the result, the weighted mean or median can also be calculated to reduce the influence of outliers. After obtaining the risk exposure of each customer, several risk classification thresholds can be set to divide the customers into different risk levels. For example: risk exposure > X is the high-risk level; X > risk exposure > Y is the medium-risk level; risk exposure < Y is the low-risk level. Then, based on these different risk levels, differentiated monitoring, warning, and control strategies can be formulated. For high-risk customers, we will preferentially allocate risk control resources and strengthen the monitoring intensity; for low-risk customers, the control intensity can be appropriately relaxed to avoid resource idleness. At the same time, by comparing the final risk exposure of different customers, we can also identify the risk area distribution characteristics within the system, such as finding that the customer groups in a certain region or industry are generally in a high-risk state, etc.

[0072] Refer to Figure 2, a credit management system based on big data risk control provided by an embodiment of the present application. The system includes: an information acquisition module, a correlation matrix determination module, a risk conduction network determination module, and a risk warning module, where: The information acquisition module is used to acquire credit information of multiple customers; The correlation matrix determination module is used to generate a lending relationship table between customers based on the lending data in each credit information, and generate a guarantee relationship table between customers based on the guarantee data in each credit information; combine the lending relationship table and the guarantee relationship table between customers to construct a customer correlation matrix; The risk conduction network determination module is used to calculate the risk factors corresponding to each customer according to the customer correlation matrix, and construct a risk conduction network according to each risk factor; The risk warning module is used to, when a preset risk event exists in the credit behavior of any customer, calculate the risk exposure of the target customer associated with the customer according to the conduction path of the customer in the risk conduction network, and generate a warning message when the risk exposure exceeds a preset threshold.

[0073] On the basis of the above embodiment, the correlation matrix determination module is further used to extract corresponding borrower information, lender information, and transaction information from each lending data; group each customer according to each lender information and borrower information to obtain lending customer groups corresponding to multiple lenders; calculate the lending intensity between each lender and the corresponding lending customer group based on each transaction information; establish a lending relationship table between each customer according to each lending intensity.

[0074] On the basis of the above embodiment, the correlation matrix determination module is further used to determine the lending times and lending amounts between each lender and the borrowers in the corresponding lending customer group within a preset time period based on each transaction information; obtain a benchmark lending intensity weight combination group, where the benchmark lending intensity weight combination group includes weight coefficients corresponding to the lending times and lending amounts; perform weighted summation on the lending times and lending amounts corresponding to each lender respectively based on the benchmark lending intensity weight combination group to obtain the lending intensity between each lender and the corresponding lending customer group.

[0075] On the basis of the above embodiment, the correlation matrix determination module is further used to determine the guarantor records and the guaranteed records of each customer based on each guarantee data; determine the guarantee chains between each customer according to each guarantee record and each guaranteed record, where the guarantee chain is a sequence of customers with a guarantee and guaranteed relationship; calculate the guarantee amount between adjacent customers in the guarantee chain; generate a guarantee relationship table between each customer according to each guarantee amount.

[0076] Based on the above embodiments, the correlation matrix determination module is further configured to obtain a group of relationship weight coefficients, where the group of relationship weight coefficients includes a lending relationship weight coefficient and a guarantee relationship weight coefficient; calculate the lending correlation degree between each customer based on the lending relationship weight coefficient and the lending intensity in the lending relationship table; calculate the guarantee correlation degree between each customer based on the guarantee relationship weight coefficient and the guarantee amount in the guarantee relationship table; generate a corresponding correlation degree vector according to the lending correlation degree and the guarantee correlation degree between each customer, and generate a customer correlation matrix according to each correlation degree vector.

[0077] Based on the above embodiments, the risk conduction network determination module is further configured to obtain the historical default records of each customer; determine the initial risk factors of each customer based on each historical default record and the customer correlation matrix; calculate the risk conduction coefficient corresponding to each customer according to a preset risk attenuation coefficient and the initial risk factors of each customer; construct a risk conduction network with each customer as a node and each risk conduction coefficient as an edge weight.

[0078] Based on the above embodiments, the risk warning module is further configured to randomly generate multiple conduction paths in the risk conduction network; obtain the occurrence frequency of the target customer in each conduction path; multiply the occurrence frequency of the target customer in each conduction path by the risk factor of the customer respectively to obtain the corresponding initial risk exposure; calculate the average value of each initial risk exposure to obtain the risk exposure corresponding to the target customer.

[0079] It should be noted that: when the device provided in the above embodiments realizes its functions, only the above-mentioned division of each functional module is used for illustration. In practical applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0080] This application also discloses an electronic device. Refer to Figure 3 , Figure 3 is a schematic structural diagram of an electronic device disclosed in an embodiment of the present application. The electronic device 300 may include: at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, and at least one communication bus 302.

[0081] Among them, the communication bus 302 is used to realize the connection and communication between these components.

[0082] Among them, the user interface 303 may include a display (Display) interface and a camera (Camera) interface. Optionally, the user interface 303 may further include a standard wired interface and a wireless interface.

[0083] Among them, the network interface 304 may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface).

[0084] Among them, the processor 301 may include one or more processing cores. The processor 301 uses various interfaces and circuits to connect various parts within the entire server. By running or executing instructions, programs, code sets, or instruction sets stored in the memory 305, and by calling the data stored in the memory 305, it executes various functions of the server and processes data. Optionally, the processor 301 may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor 301 may integrate a combination of one or more of a central processing unit (CPU), a graphics processing unit (GPU), and a modem, etc. Among them, the CPU mainly processes the operating system, user interface graphics, and application programs, etc.; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 301 and may be implemented separately by a single chip.

[0085] Among them, the memory 305 may include random access memory (RAM) and may also include read-only memory. Optionally, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing the operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store the data involved in the above-mentioned various method embodiments. The memory 305 is optionally also at least one storage device located far from the aforementioned processor 301. Refer to Figure 3 , the memory 305, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program of a credit management method based on big data risk control.

[0086] In Figure 3 In the electronic device 300 shown, the user interface 303 is mainly used to provide an interface for the user to input and obtain the data input by the user; and the processor 301 can be used to call the application program stored in the memory 305 for a credit management method based on big data risk control. When executed by one or more processors 301, the electronic device 300 is caused to execute the method as described in one or more of the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be adopted in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0087] In the above embodiments, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0088] In several implementation manners provided by this application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some service interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical or other form.

[0089] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0090] In addition, in each embodiment of this application, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0091] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present application. The aforementioned memory includes various media that can store program codes, such as USB flash drives, mobile hard disks, magnetic disks, or optical discs.

[0092] The above are only exemplary embodiments of the present disclosure, and the scope of the present disclosure cannot be limited thereby. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure still fall within the scope covered by the present disclosure. Those skilled in the art will readily think of other embodiments of the present disclosure after considering the specification and the practice of the disclosure.

[0093] The present application aims to cover any variations, uses, or adaptive changes of the present disclosure. These variations, uses, or adaptive changes follow the general principles of the present disclosure and include common general knowledge or conventional technical means in the technical field not recorded in the present disclosure. The specification and the embodiments are only regarded as exemplary.

Claims

1. A credit management method based on big data risk control, characterized in that: include: Obtain credit information for multiple customers; Based on the loan data in each of the credit information, a loan relationship table between each of the customers is generated, and based on the guarantee data in each of the credit information, a guarantee relationship table between each of the customers is generated; Combine the loan relationship table and guarantee relationship table between the customers to build a customer correlation matrix; Calculating the risk factor corresponding to each of the customers according to the customer association matrix, and constructing a risk transmission network according to each of the risk factors; When there is a preset risk event in the credit behavior of any customer, the risk exposure corresponding to the target customer associated with the customer is calculated according to the transmission path of the customer in the risk transmission network, and warning information is generated when the risk exposure exceeds the preset threshold.

2. The credit management method based on big data risk control according to claim 1 is characterized in that: The generating of a loan relationship table between each of the customers based on the loan data in each of the credit information includes: Extracting corresponding borrower information, lender information and transaction information from each of the loan data; Grouping the customers according to the lender information and borrower information to obtain borrower customer groups corresponding to multiple lenders; Based on each of the transaction information, calculating the loan intensity between each of the lenders and the corresponding loan customer group; A lending relationship table between the customers is established based on the lending strengths.

3. The credit management method based on big data risk control according to claim 2 is characterized in that: The calculating, based on each of the transaction information, the loan intensity between each of the lenders and the corresponding loan customer group comprises: Based on each of the transaction information, determine the number of loans and loan amounts between each of the lenders and the borrowers in the corresponding lending customer group within a preset time period; Obtaining a benchmark lending strength weight group, wherein the benchmark lending strength weight group includes weight coefficients corresponding to the number of loans and the loan amount; Based on the benchmark lending intensity weight group, the number of loans and the amount of loans corresponding to each lender are weighted and summed to obtain the lending intensity between each lender and the corresponding lending customer group.

4. The credit management method based on big data risk control according to claim 1 is characterized in that: The generating of a guarantee relationship table between the customers based on the guarantee data in the credit information includes: Based on each of the guarantee data, determining a guarantor record and a guaranteed person record for each of the customers; Determine the guarantee chain between the customers according to the guarantee records and the guaranteed person records, wherein the guarantee chain is a sequence of customers having a guarantee and guaranteed relationship; Calculating the guarantee amounts between adjacent customers in the guarantee chain; A guarantee relationship table between the customers is generated based on the guarantee amounts.

5. The credit management method based on big data risk control according to claim 1, characterized in that: The step of combining the loan relationship table and the guarantee relationship table between the customers to construct a customer association matrix includes: Acquire a relationship weight coefficient group, wherein the relationship weight coefficient group includes a loan relationship weight coefficient and a guarantee relationship weight coefficient; Calculating the degree of loan association between each of the customers based on the loan relationship weight coefficient and the loan intensity in the loan relationship table; Calculate the guarantee association between the customers based on the guarantee relationship weight coefficient and the guarantee amount in the guarantee relationship table; According to the loan correlation and guarantee correlation between the customers, a corresponding correlation vector is generated, and according to the correlation vectors, a customer correlation matrix is ​​generated.

6. The credit management method based on big data risk control according to claim 1, characterized in that: The step of calculating the risk factor corresponding to each of the customers according to the customer association matrix, and constructing a risk transmission network according to each of the risk factors, includes: Obtaining historical default records of each of the customers; Determining an initial risk factor for each of the customers based on each of the historical default records and the customer association matrix; Calculate the risk transmission coefficient corresponding to each of the customers according to the preset risk attenuation coefficient and the initial risk factor of each of the customers; A risk transmission network is constructed with each of the customers as a node and each of the risk transmission coefficients as an edge weight.

7. The credit management method based on big data risk control according to claim 1 is characterized in that: The calculating, according to the conduction path of the customer in the risk conduction network, the risk exposure corresponding to the target customer associated with the customer comprises: randomly generating a plurality of conduction paths in the risk conduction network; Obtaining the appearance frequency of the target customer in each of the transmission paths; Multiplying the frequency of occurrence of the target customer in each of the transmission paths by the risk factor of the customer to obtain the corresponding initial risk exposure; An average value of each of the initial risk exposures is calculated to obtain a risk exposure corresponding to the target customer.

8. A credit management system based on big data risk control, characterized in that: The system comprises: An information acquisition module is used to obtain credit information of multiple customers; A correlation matrix determination module is used to generate a loan relationship table between each of the customers based on the loan data in each of the credit information, and to generate a guarantee relationship table between each of the customers based on the guarantee data in each of the credit information; and to construct a customer correlation matrix by combining the loan relationship table and the guarantee relationship table between each of the customers; A risk transmission network determination module is used to calculate the risk factor corresponding to each of the customers according to the customer association matrix, and to construct a risk transmission network according to each of the risk factors; The risk warning module is used to calculate the risk exposure corresponding to the target customers associated with any customer based on the conduction path of the customer in the risk conduction network when there is a preset risk event in the credit behavior of any customer, and generate warning information when the risk exposure exceeds a preset threshold.

9. An electronic device, characterized in that: It includes a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the credit management method based on big data risk control as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed, the credit management method based on big data risk control as described in any one of claims 1 to 7 is executed.

Images