A bank credit risk management method based on big data analysis

Abstract

The application relates to a bank credit risk management method based on big data analysis, and belongs to the technical field of credit risk management. The method comprises the following steps: receiving a risk assessment instruction, collecting multi-source heterogeneous data of a credit user corresponding to the risk assessment instruction; analyzing and integrating the multi-source heterogeneous data, generating and outputting a credit user portrait; obtaining an intervention condition corresponding to the user portrait, deducing and analyzing an intervention result of the intervention condition on the credit user through a pre-constructed counterfactual risk deduction model, and outputting the intervention result for bank credit management personnel to know; wherein the intervention condition refers to a market situation that has an impact on the credit default risk of a credit user; and the intervention result at least includes a credit default probability. The application has the effect of optimizing the risk prediction ability of a credit user in a dynamic market situation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of credit risk management, in particular to a bank credit risk management method based on big data analysis. BACKGROUND

[0002] With the rapid development of financial technology, banks are also facing more and more challenges in credit risk management. In recent years, the rise of big data technology has provided a new solution for credit risk management by collecting and analyzing massive data of credit users to analyze the risk status of credit users.

[0003] However, the traditional credit risk assessment method mainly relies on historical data and static models, and when facing dynamic and changing market environment, the aforementioned analysis technology has limited risk prediction ability of credit users in dynamic market situation, and therefore needs to be improved. SUMMARY

[0004] In order to optimize the risk prediction ability of credit users in dynamic market situation, the present application provides a bank credit risk management method based on big data analysis.

[0005] In a first aspect, the present application provides a bank credit risk management method based on big data analysis, comprising:

[0006] receiving a risk assessment instruction, collecting multi-source heterogeneous data of a credit user corresponding to the risk assessment instruction;

[0007] analyzing and integrating the multi-source heterogeneous data to generate and output a credit user portrait;

[0008] obtaining an intervention condition corresponding to the user portrait, deducing and analyzing the intervention result of the intervention condition on the credit user through a pre-constructed counterfactual risk deduction model, and outputting the intervention result for bank credit management personnel to know; wherein the intervention condition refers to a market situation that has an impact on the credit default risk of a credit user; the intervention result at least includes a credit default probability.

[0009] By adopting the above technical solution, multi-source heterogeneous data of a credit user is collected to help more comprehensively analyze and obtain a credit user portrait, and the development (such as production and operation behavior) of a credit user is intervened by deducing and analyzing the intervention condition that may occur in the future, and the influence degree of the intervention condition on the credit user (and especially the influence situation on the credit default probability of the credit user) is deduced and analyzed, so as to help the bank to predict the enterprise default risk and enhance the forward-looking and stress testing ability of risk control.

[0010] Optionally, the intervention condition includes an initial intervention condition and a derived intervention condition.

[0011] The intervention condition corresponding to the user portrait is obtained, including:

[0012] An initial intervention condition corresponding to the user portrait is obtained, and a derivative intervention condition is generated by analyzing the initial intervention condition through a preset GAN risk simulation model; and the initial intervention condition and the derivative intervention condition are combined as the intervention condition.

[0013] By adopting the above technical solution, the initial intervention condition can be an intervention condition set by the bank based on the user portrait and the explicit risk direction that the credit user is likely to experience, and the application further automatically generates unknown market scenarios by using the GAN risk simulation model, thereby increasing the sample size of the market scenarios (i.e., the number of intervention conditions) to more comprehensively assist the counterfactual risk deduction model in analyzing the risk impact that the credit user may suffer, helping to discover hidden risks, further enhancing the forward-looking and stress testing capabilities of risk control, and the comprehensiveness of the prediction of the default risk of the credit user.

[0014] Optionally, the analysis and integration of the multi-source heterogeneous data to generate and output a credit user portrait includes:

[0015] The multi-source heterogeneous data is analyzed, and sensitive indicators are extracted from the multi-source heterogeneous data based on a preset sensitivity analysis algorithm; wherein the sensitive indicators are data indicators that have an impact on credit default risk;

[0016] Based on the multi-source heterogeneous data, an association graph of the credit user is determined, and the association graph at least includes an industry association graph;

[0017] A credit user portrait with sensitive indicators and an industry association graph is generated and output;

[0018] The initial intervention condition corresponding to the user portrait is obtained, including:

[0019] A first initial intervention condition is generated based on the sensitive indicators;

[0020] Based on the association graph, a risk association transmission path is determined, and a second initial intervention condition is generated according to the risk association transmission path;

[0021] A third initial intervention condition is obtained in real time by quantifying macroeconomic events from the macroeconomic events;

[0022] The first initial intervention condition, the second initial intervention condition, and the third initial intervention condition are integrated to obtain the initial intervention condition.

[0023] By adopting the technical scheme, the sensitivity of the credit user is analyzed, the associated risk conduction path is analyzed to obtain the joint risk, and the risk existing in the macroeconomic event is extracted in real time, so as to analyze the intervention condition from multiple dimensions, and the intervention condition is automatically generated to replace the manual setting of the intervention condition, so that the dynamic risk assessment is more intelligent and efficient.

[0024] Optionally, the method further comprises:

[0025] The risk source is obtained, and the risk associated conduction path corresponding to the risk source is determined based on the association graph, and the risk associated conduction path is: the conduction path of the risk source causing the credit default risk of the credit user when the risk source is conducted according to the association relationship in the association graph;

[0026] The risk conduction intensity of the determined risk associated conduction path is calculated and determined, wherein the risk conduction intensity is used to represent the influence degree of the credit default influence of the credit user caused by the risk source when the risk source is conducted according to the corresponding risk associated conduction path;

[0027] The second initial intervention condition is generated according to the risk associated conduction path, and the corresponding relationship of the second initial intervention condition, the corresponding risk conduction intensity, and the corresponding risk associated conduction path is output, so that the bank credit management personnel can know.

[0028] By adopting the technical scheme, the influence degree of the credit default risk of the credit user caused by the risk source is visualized and displayed by using the association graph, and the influence degree is quantified as the risk conduction intensity, which helps the bank credit management personnel to more intuitively know all possible risk shocks and the risk shock degree of the credit user by the risk associated conduction path.

[0029] Optionally, the method further comprises:

[0030] Based on the preset analysis algorithm, the occurrence probability of the intervention condition and the analysis basis corresponding to the occurrence probability are analyzed, and the corresponding relationship of the intervention condition, the occurrence probability and the corresponding analysis basis is output.

[0031] By adopting the technical scheme, the occurrence probability of the intervention condition is analyzed, and the analysis basis is output, so as to avoid the over-protection behavior caused by the credit default risk of the low-probability intervention condition, and the resource allocation can be optimized based on the probability, such as the credit default risk caused by the high-probability intervention condition.

[0032] Optionally, the method further comprises:

[0033] Before outputting the intervention result, the intervention result is sent to an expert review terminal, so that an expert reviews and approves the intervention result through the expert review terminal, and the intervention result is updated based on the expert review result.

[0034] By adopting the technical solution, in order to avoid the occurrence of extremely rare situations of the automatically generated intervention condition, the application further proposes a manual intervention review scheme to optimize the objective rationality of the credit risk assessment of the credit user.

[0035] Optionally, the outputting the intervention result comprises:

[0036] The intervention condition and the corresponding intervention result are outputted in a preset display form based on a preset visual explanation module; wherein the preset display form at least includes a chart form.

[0037] By adopting the technical solution, the intervention condition and the intervention result are directly displayed in the form of icons and the like, so as to assist the bank credit management personnel to efficiently understand the influence process of the intervention condition on the credit risk of the credit user.

[0038] In a second aspect, the application provides a bank credit risk management system based on big data analysis, comprising,

[0039] A multi-source data acquisition module is configured to receive a risk assessment instruction and collect multi-source heterogeneous data of a credit user corresponding to the risk assessment instruction.

[0040] A user portrait generation module is configured to analyze and integrate the multi-source heterogeneous data, generate and output a credit user portrait.

[0041] A risk simulation and deduction module is configured to obtain an intervention condition corresponding to the user portrait, deduce and analyze an intervention result of the credit user under the intervention condition through a pre-constructed counterfactual risk deduction model, and output the intervention result for a bank credit management personnel to know; wherein the intervention condition refers to a market situation that has an impact on the credit default risk of the credit user; and the intervention result at least includes a credit default probability.

[0042] In a third aspect, the application provides a bank credit risk management device based on big data analysis, comprising a memory and a processor, and the memory stores a computer program capable of being loaded and executed by the processor and performing the method of any one of the first aspect.

[0043] In a fourth aspect, the application provides a computer readable storage medium, which stores a computer program capable of being loaded and executed by a processor and performing the method of any one of the first aspect.

[0044] In summary, the application has the following beneficial technical effects:

[0045] In the present application, by collecting multi-source heterogeneous data of credit users, a credit user portrait is comprehensively analyzed, and the development of the credit user is intervened by using future possible intervention conditions, the influence degree of the intervention conditions on the credit user is obtained by deduction analysis, so as to help the bank to predict the enterprise default risk and enhance the forward-looking and stress testing ability of risk control. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0047] Figure 1 is a flowchart of the bank credit risk management method based on big data analysis of the embodiments of the present application.

[0048] Figure 2 is a structural block diagram of the bank credit risk management system based on big data analysis of the embodiments of the present application.

[0049] Marked with 201, multi-source data acquisition module; 202, user portrait generation module; 203, risk simulation deduction module. DETAILED DESCRIPTION

[0050] The following will be described in conjunction with the accompanying Figures 1-2 The present application will be further described in detail.

[0051] The embodiments of the present application disclose a bank credit risk management method based on big data analysis (hereinafter referred to as credit risk management method), which is used to collect and analyze multi-source data (such as financial data) of credit users, and to predict the credit default risk of enterprises under different market mutation situations based on multi-source data, and to optimize the risk prediction ability of credit users under dynamic market situation. The execution subject of the credit risk management method is a bank credit risk management system based on big data analysis (hereinafter referred to as credit risk management system), which will be described in conjunction with the accompanying Figure 1 The execution steps of the credit risk management system on the credit risk management method will be described in detail.

[0052] S101, receiving a risk assessment instruction, collecting multi-source heterogeneous data of a credit user corresponding to the risk assessment instruction.

[0053] In implementation, the risk assessment instruction can be triggered by a bank credit manager when accessing the credit risk management system, wherein the risk assessment instruction at least contains a credit user and uploaded multi-source heterogeneous data related to the credit user. The credit user of the present application refers to a user who applies for a loan to a bank, which can be an individual or an enterprise. In the embodiments of the present application, an enterprise is taken as a credit user for further elaboration. The multi-source heterogeneous data of the credit user can specifically include enterprise fundamental data (such as business registration information, financial data, judicial information), operating behavior data (such as bank flow, tax data, supply chain data), industry and market data (industry sentiment index (such as PMI, capacity utilization rate, price index), public opinion data, competition pattern), enterprise correlation network data (such as guarantee circle data, equity control chain), unstructured data (such as financial report appendix text (such as accounting policy change or liability disclosure), customer evaluation (such as supplier / customer complaint, cooperation stability evaluation), on-site investigation data (such as factory equipment status, employee interview record)) and the like. The foregoing data can be collected and uploaded by the credit user himself, or can be searched by the credit risk management system, such as obtaining public opinion data, industry research report / e-commerce platform data through web crawler; obtaining tax data through tax bureau authorized interface, obtaining industry sentiment index through national statistical bureau / industry association authorized interface, and the like. It should be noted that the above multi-source heterogeneous data are obtained in a legal way under the condition that the credit user knows and authorizes, and the data source platform is authorized.

[0054] S102, analyzing and integrating the multi-source heterogeneous data, generating and outputting a credit user portrait;

[0055] S102 specifically includes the following contents:

[0056] analyzing the multi-source heterogeneous data, and extracting sensitive indicators from the multi-source heterogeneous data based on a preset sensitivity analysis algorithm; wherein the sensitive indicators are data indicators that have an impact on credit default risk;

[0057] determining a correlation graph of the credit user based on the multi-source heterogeneous data, wherein the correlation graph at least includes an industry correlation graph;

[0058] generating and outputting a credit user portrait with sensitive indicators and an industry correlation graph.

[0059] In implementation, the credit risk management system is used to analyze the financial data (such as debt ratio, customer concentration) and operating data (such as supply chain length, inventory turnover) in the multi-source heterogeneous data of the enterprise, and the sensitivity test is performed through the preset sensitivity analysis algorithm (such as gradient sensitivity calculation), and the sensitive indicators of the enterprise are screened out (for example, a lightweight model such as a linear regression model can be trained using historical data of the enterprise to calculate the partial derivative of each indicator on the credit default probability, and the sensitive variable is found out, and the indicator corresponding to the sensitive variable is the sensitive indicator), such as "if the cost of raw materials rises, the enterprise's profit margin decreases by 30% higher than the industry average", and the corresponding sensitive indicator determined is the raw material price.

[0060] The association graph of the credit user refers to the enterprise relationship graph, that is, the relationship network formed by the direct / indirect associated parties (such as suppliers, customers, guarantee circle) of the enterprise, and the credit risk management system can be based on the enterprise association network data in the multi-source heterogeneous data, such as business equity data (parent-subsidiary companies, actual controllers), supply chain data (upstream suppliers / downstream customer list), guarantee circle data (mutual guarantee, joint guarantee relationship), etc. For example, the association graph can specifically include a guarantee relationship network, which can be constructed in the following form: enterprise A guarantees enterprise B, and enterprise B supplies goods to enterprise C.

[0061] Finally, the credit user portrait describing the specific situation of the credit user is obtained, and the association graph and the sensitive indicator obtained above are specific forms of the credit user portrait.

[0062] S103, obtaining an intervention condition corresponding to the user portrait, deducing and analyzing the intervention result of the intervention condition on the credit user through a pre-constructed counterfactual risk deduction model, and outputting the intervention result for the bank credit management personnel to know; wherein the intervention condition refers to a market situation that has an impact on the credit default risk of the credit user; the intervention result at least includes the credit default probability.

[0063] In S103, "obtaining an intervention condition corresponding to the user portrait" specifically includes:

[0064] S1031, generating a first initial intervention condition based on the sensitive indicator;

[0065] S1032, determining a risk association transmission path based on the association graph, and generating a second initial intervention condition according to the risk association transmission path;

[0066] S1033, real-time acquisition of macroeconomic events, and quantification of a third initial intervention condition from the macroeconomic events;

[0067] S1034, integrating the first initial intervention condition, the second initial intervention condition, and the second initial intervention condition to obtain an initial intervention condition;

[0068] S1035, obtaining an initial intervention condition corresponding to the user portrait, analyzing the initial intervention condition through a preset GAN risk simulation model to generate a derived intervention condition; and merging the initial intervention condition and the derived intervention condition as the intervention condition.

[0069] S1032 specifically includes the following contents:

[0070] obtaining a risk source, determining a risk association conduction path corresponding to the risk source based on the association graph, and the risk association conduction path is a conduction path of the risk source causing a credit default risk to the credit user when the risk source is conducted according to the association relationship in the association graph;

[0071] calculating and determining the risk conduction intensity of the risk association conduction path, wherein the risk conduction intensity is used to represent the influence degree of the risk source causing the credit default influence to the credit user when the risk source is conducted according to the corresponding risk association conduction path;

[0072] generating a second initial intervention condition according to the risk association conduction path, outputting the second initial intervention condition and the corresponding risk conduction intensity, and the corresponding relationship of the corresponding risk association conduction path, so as to be known by the bank credit management personnel.

[0073] In implementation, the state of each node (an associated party associated with the credit user) in the association graph is monitored in real time, and the public opinion information is monitored and obtained in real time, and each time the public opinion information is obtained, it is determined whether the public opinion information is associated with any node in the association graph. For example, the public opinion information is: the news reports that "Australia's lithium mine exports have decreased by 40%", find the node associated with the public opinion information and mark it as a risk source, such as determining the node associated with the lithium mine, such as the decrease in lithium mine exports will affect the price increase of lithium carbonate, which in turn will affect the production and operation of lithium batteries. The corresponding associated node can be a lithium battery factory, an electric vehicle related industry, etc.

[0074] After determining the risk source, all risk association conduction paths (such as all risk association conduction paths can be found based on breadth-first search BFS or random walk search technology) are found, and the risk association conduction path satisfies: the path contains the node corresponding to the credit user and the node corresponding to the risk source. In other words, the risk association conduction path refers to all conduction paths of the risk source to the credit user. For example, in combination with the above example, if the credit user is an electric vehicle manufacturer, then the corresponding risk association conduction path can include:

[0075] Path 1: Price increase of lithium carbonate → cost increase of lithium battery → price increase of electric vehicle → sales decrease of credit user

[0076] Path 2: Price increase of lithium carbonate → profit decrease of lithium battery factory → credit user reduces R&D investment → technology development lags

[0077] Path 3: Lithium carbonate price rise -> Small battery factory bankruptcy -> Supply chain disruption -> Electric vehicle production decline.

[0078] Next, the credit risk management system will further quantify the risk transmission intensity of each risk associated transmission path, that is, the degree of influence on the credit default probability of the credit user. The present application proposes that in the process of calculating the risk transmission intensity, three dimensions are considered: industry dependence D, transmission delay L, and adjustment factor M (defined as a value between 0 and 1). Industry dependence refers to the degree of dependence of the credit user on the previous transmission node in the corresponding path. Small battery factory A is a supplier of the credit user, and the credit user's procurement from small battery factory A accounts for X% of the total procurement. Compared to the industry average procurement ratio Y, the dependence D = X / Y. Transmission delay is used to quantify the transmission time lag of the risk according to the corresponding path. Specifically, the Granger causality test method is used to calculate the transmission time based on the historical transmission time (such as the 2-month transmission time from lithium carbonate price rise to electric vehicle price rise). The adjustment factor refers to policy buffer (such as export tax rebate offsetting part of the cost increase of lithium battery) or enterprise risk resistance ability (such as inventory turnover days). For example, the risk transmission intensity calculation formula can be S = D * e -βL * M. Thus, the risk transmission intensity of each risk associated transmission path is obtained.

[0079] Next, the credit risk management system will generate intervention conditions, which include initial intervention conditions and derived intervention conditions. The initial intervention conditions include first initial intervention conditions, second initial intervention conditions, and third initial intervention conditions. The derived intervention conditions are further refined by the GAN risk simulation model based on the initial intervention conditions.

[0080] Specifically, the first initial intervention condition is generated based on the sensitive index. After randomly defining the value of the sensitive index, the sensitive first initial intervention condition is generated. For example, if the sensitive index is the raw material price, then the first initial intervention condition generated is "raw material price + 15%".

[0081] The second initial intervention condition is generated based on each node in the risk associated transmission path. For example, for the "lithium carbonate price rise" node, the second intervention condition "lithium carbonate price + 50%" is automatically generated. For the "lithium battery factory profit decline" node, the second intervention condition "lithium battery factory profit - 25%" is automatically generated.

[0082] The third initial intervention condition is an intervention condition generated by the credit risk management system from unstructured events (such as news) through a preset extraction model (such as a BERT model, an NPL event extraction) after keywords are extracted. For example, if the news is “the United States announces restrictions on semiconductor exports to China”, the extracted keywords are “semiconductor imports”, and the corresponding intervention condition can be “semiconductor import cost + 30%”.

[0083] After obtaining the initial intervention condition, the GAN risk simulation model is further refined on the basis of the initial intervention condition. For example, the initial intervention condition “lithium battery price + 50%” is further derived into “lithium battery price + 50% for 6 months”, that is, the derived intervention condition is obtained by adding additional conditions to the initial intervention condition. Then, the counterfactual risk deduction model is used to deduce and analyze the intervention result of each intervention condition on the credit user. The intervention result at least includes the credit risk default probability, such as the credit risk default probability from 5% to 22%.

[0084] Optionally, the credit risk management method further includes the following steps:

[0085] Based on the preset analysis algorithm, the occurrence probability of the intervention condition is analyzed, and the analysis basis corresponding to the occurrence probability is analyzed. The corresponding relationship between the intervention condition, the occurrence probability, and the corresponding analysis basis is output.

[0086] In implementation, the historical data statistical method can be used to count the occurrence frequency of each intervention condition in the historical period, which is used as the occurrence probability of the intervention condition. The corresponding analysis basis is the number of examples of the corresponding intervention condition in the historical period. In addition, the occurrence probability of the intervention condition (such as policy mutation) influenced by complex factors can also be analyzed by machine learning prediction method. For example, the classification model (such as XGBoost) is used to predict whether the intervention condition will occur, and the probability model (such as Bayesian network) is used to calculate the occurrence probability. In other embodiments, the occurrence probability of each intervention condition can also be determined by manual intervention. Correspondingly, the credit risk management method further includes the following steps:

[0087] Before outputting the intervention result, the intervention result is delivered to an expert review terminal, so that the expert reviews and approves the intervention result through the expert review terminal, and updates the intervention result based on the expert review result.

[0088] All intervention conditions and corresponding intervention structures are delivered to the expert review terminal, and the expert defines the occurrence probability of each intervention condition in the form of annotations and annotates the analysis basis (such as artificial experience or attached artificial judgment basis), and finally obtains the intervention result with expert annotations.

[0089] Optionally, when outputting the intervention result, the application further proposes: based on a preset visual explanation module, outputting the intervention condition and the corresponding intervention result in a preset display form; wherein the preset display form at least includes a chart form. That is, the corresponding relationship between the risk associated transmission path, the intervention condition and the corresponding intervention result is displayed in various forms. For example, the risk associated transmission path is displayed in a chart form of a force transmission diagram + a heat map layer. Different colors are used to distinguish the risk source and the affected node, arrows (such as “→”) are used to represent the transmission direction in the risk associated transmission path, and the arrow thickness is used to represent the transmission strength of the corresponding path. For example, the specific content of different intervention conditions, corresponding occurrence probabilities and intervention results is displayed in a table form.

[0090] The application also discloses a bank credit risk management system based on big data analysis. Referring to Figure 2 , comprising:

[0091] The multi-source data acquisition module 201 is configured to receive a risk assessment instruction, and collect multi-source heterogeneous data of a credit user corresponding to the risk assessment instruction.

[0092] The user portrait generation module 202 is configured to analyze and integrate the multi-source heterogeneous data, and generate and output a credit user portrait.

[0093] The risk simulation and deduction module 203 is configured to obtain an intervention condition corresponding to the user portrait, analyze an intervention result of the intervention condition on the credit user through a pre-constructed counterfactual risk deduction model, and output the intervention result for a bank credit management personnel to know. The intervention condition refers to a market situation that has an impact on the credit default risk of a credit user. The intervention result at least includes a credit default probability.

[0094] Optionally, the risk simulation and deduction module 203 is further configured to obtain an initial intervention condition corresponding to the user portrait, generate a derived intervention condition by analyzing the initial intervention condition through a preset GAN risk simulation model, and combine the initial intervention condition and the derived intervention condition as the intervention condition.

[0095] Optionally, the user portrait generation module 202 is further configured to analyze the multi-source heterogeneous data, extract a sensitive index from the multi-source heterogeneous data based on a preset sensitivity analysis algorithm, and determine an association graph of the credit user based on the multi-source heterogeneous data. The association graph at least includes an industry association graph. The credit user portrait with the sensitive index and the industry association graph is generated and output.

[0096] The risk simulation deduction module 203 is further configured to generate a first initial intervention condition based on the sensitive indicators; determine a risk correlation transmission path based on the correlation graph, and generate a second initial intervention condition according to the risk correlation transmission path; acquire macroeconomic events in real time, and quantitatively obtain a third initial intervention condition from the macroeconomic events; and integrate the first initial intervention condition, the second initial intervention condition, and the third initial intervention condition to obtain an initial intervention condition.

[0097] Optionally, the risk simulation deduction module 203 is further configured to acquire a risk source, determine a risk correlation transmission path corresponding to the risk source based on the correlation graph, and the risk correlation transmission path is a transmission path of the risk source causing a credit default risk of a credit user when the risk source is transmitted according to the correlation relationship in the correlation graph; calculate and determine a risk transmission intensity of the determined risk correlation transmission path, wherein the risk transmission intensity is used to represent an influence degree of the risk source causing the credit default risk of the credit user when the risk source is transmitted according to the corresponding risk correlation transmission path; generate a second initial intervention condition according to the risk correlation transmission path, and output a corresponding relationship between the second initial intervention condition and the corresponding risk transmission intensity and the corresponding risk correlation transmission path, so as to be known by a bank credit management personnel.

[0098] Optionally, the device further comprises a probability analysis module configured to analyze and obtain an occurrence probability of the intervention condition and an analysis basis corresponding to the occurrence probability based on a preset analysis algorithm, and output a corresponding relationship between the intervention condition, the occurrence probability, and the analysis basis.

[0099] Optionally, the device further comprises an expert review module configured to, before outputting the intervention result, transmit the intervention result to an expert review terminal, so that an expert reviews and approves the intervention result through the expert review terminal, and updates the intervention result based on an expert review result.

[0100] Optionally, the risk simulation deduction module 203 is further configured to output the intervention condition and the corresponding intervention result in a preset display form based on a preset visual explanation module; wherein the preset display form at least includes a chart form.

[0101] The embodiment of the application further discloses a bank credit risk management device based on big data analysis, which comprises a memory and a processor, and the memory stores a computer program capable of being loaded and executed by the processor to execute the bank credit risk management method based on big data analysis.

[0102] The embodiment of the application further discloses a computer readable storage medium which stores a computer program capable of being loaded by a processor and executing the bank credit risk management method based on big data analysis as described above, and the computer readable storage medium comprises various storage medium capable of storing program codes, for example, a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0103] It should be noted that, in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations.

[0104] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit the protection scope of the application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on these embodiments, all other embodiments obtained by those of ordinary skill in the art without creative labor shall fall within the scope of protection of the present application.

Claims

1. A big data analysis-based bank credit risk management method, characterized by, The method comprises the following steps: receiving a risk assessment instruction, collecting multi-source heterogeneous data of a credit user corresponding to the risk assessment instruction; analyzing and integrating the multi-source heterogeneous data to generate and output a credit user portrait; obtaining an intervention condition corresponding to the user portrait, deducing and analyzing the intervention result of the intervention condition on the credit user through a pre-constructed counterfactual risk deduction model, and outputting the intervention result for a bank credit manager to know; wherein the intervention condition refers to a market situation that has an impact on the credit default risk of a credit user; the intervention result at least includes a credit default probability; the intervention condition includes an initial intervention condition and a derived intervention condition; the obtaining of the intervention condition corresponding to the user portrait comprises: obtaining an initial intervention condition corresponding to the user portrait, generating a derived intervention condition after analyzing the initial intervention condition through a preset GAN risk simulation model; and combining the initial intervention condition and the derived intervention condition as the intervention condition; the analysis and integration of the multi-source heterogeneous data to generate and output a credit user portrait comprises: analyzing the multi-source heterogeneous data, and extracting sensitive indicators from the multi-source heterogeneous data based on a preset sensitivity analysis algorithm; wherein the sensitive indicators refer to data indicators that have an impact on the credit default risk; determining an association graph of the credit user based on the multi-source heterogeneous data, wherein the association graph at least includes an industry association graph; generating and outputting a credit user portrait with sensitive indicators and an industry association graph; the obtaining of the initial intervention condition corresponding to the user portrait comprises: generating a first initial intervention condition based on the sensitive indicators; determining a risk association transmission path based on the association graph, and generating a second initial intervention condition according to the risk association transmission path; real-time obtaining of macroeconomic events, and quantitatively obtaining a third initial intervention condition from the macroeconomic events; integrating the first initial intervention condition, the second initial intervention condition and the third initial intervention condition to obtain the initial intervention condition.

2. The big data analysis based bank credit risk management method according to claim 1, characterized in that, the determination of the risk association transmission path based on the association graph and the generation of the second initial intervention condition according to the risk association transmission path comprises: obtaining a risk source, determining a risk association transmission path corresponding to the risk source based on the association graph, and the risk association transmission path is: the transmission path of the risk source causing the credit default risk of the credit user when the risk source transmits according to the association relationship in the association graph; calculating and determining the risk transmission intensity of the obtained risk association transmission path, wherein the risk transmission intensity is used to represent the influence degree of the risk source on the credit default of the credit user when the risk source transmits according to the corresponding risk association transmission path; generating a second initial intervention condition according to the risk association transmission path, outputting the corresponding relationship between the second initial intervention condition and the corresponding risk transmission intensity, and the corresponding risk association transmission path, for a bank credit manager to know. 3.The big data analysis based bank credit risk management method according to claim 1, characterized in that, the method further comprises: Based on a preset analysis algorithm, an occurrence probability of the intervention condition is analyzed, and an analysis basis corresponding to the occurrence probability is obtained; and a corresponding relationship between the intervention condition, the occurrence probability, and the corresponding analysis basis is output. 4.The big data analysis based bank credit risk management method according to claim 1, characterized in that, The method further includes: Before outputting the intervention result, the intervention result is delivered to an expert review terminal, so that an expert reviews and approves the intervention result through the expert review terminal, and the intervention result is updated based on an expert review result. 5.The big data analysis based bank credit risk management method according to claim 1, characterized in that, The outputting of the intervention result includes: Based on a preset visual interpretation module, the intervention condition and the corresponding intervention result are output in a preset display form; wherein the preset display form at least includes a chart form. 6.A bank credit risk management system based on big data analysis, applied to the bank credit risk management method based on big data analysis of claim 1, characterized in that, The bank credit risk management system based on big data analysis includes: A multi-source data acquisition module (201) is configured to receive a risk assessment instruction and collect multi-source heterogeneous data of a credit user corresponding to the risk assessment instruction; A user portrait generation module (202) is configured to analyze and integrate the multi-source heterogeneous data, generate and output a credit user portrait; A risk simulation and deduction module (203) is configured to obtain an intervention condition corresponding to the user portrait, deduce and analyze an intervention result of the credit user by the intervention condition through a pre-constructed counterfactual risk deduction model, and output the intervention result for a bank credit management personnel to know; wherein the intervention condition refers to a market situation that has an impact on a credit default risk of a credit user; and the intervention result at least includes a credit default probability. 7.A big data analysis based bank credit risk management device, characterized by, A memory and a processor are included, and the memory stores a computer program capable of being loaded and executed by the processor to perform the method of any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, A memory and a processor are included, and the memory stores a computer program capable of being loaded and executed by the processor to perform the method of any one of claims 1 to 5.

Images