Customer credit risk prediction method, device, equipment, medium and program product

Abstract

The invention provides a customer credit risk prediction method which can be applied to the technical field of artificial intelligence. The customer credit risk prediction method comprises the steps of obtaining multiple groups of feature information based on a customer type of a target customer; wherein the multiple groups of feature information are obtained based on the original information of the target customer, and the original information corresponding to different customer types is different; selecting a target mode of a prediction model based on the customer type and the feature information; wherein the prediction model is a multi-mode model, and parameters of the prediction model in different modes are different; inputting the multiple groups of feature information into a prediction model, and analyzing the multiple groups of feature information by the prediction model based on the target mode to obtain multiple pieces of first information; wherein the first information is used for predicting performance conditions of the target customer in different dimensions; and determining the credit risk of the target customer based on the multiple pieces of first information. The invention further provides a customer credit risk prediction device and equipment, a storage medium and a program product.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and more specifically to a customer credit risk prediction method, device, equipment, medium and program product. Background Art

[0002] Customer credit risk prediction can help banks and other financial institutions understand customers' financial capabilities in a timely manner, reduce transaction risks and minimize losses.

[0003] Existing customer credit risk prediction usually adopts a unified method to predict customer credit risk, ignoring the differences between different types of customers and failing to flexibly conduct specific analysis and judgment based on the situations of different customers, resulting in a low accuracy rate of credit risk prediction results. Summary of the invention

[0004] In view of the above problems, the present disclosure provides a customer credit risk prediction, device, equipment, medium and program product for improving the accuracy of customer credit risk prediction.

[0005] According to a first aspect of the present disclosure, a method for predicting customer credit risk is provided, comprising: obtaining multiple groups of feature information based on customer types of target customers; wherein the multiple groups of feature information are obtained based on original information of the target customers, and different original information corresponds to different customer types; selecting a target mode of a prediction model based on the customer type and the feature information; wherein the prediction model is a multi-mode model, and the parameters of the prediction model are different in different modes; inputting the multiple groups of feature information into the prediction model, and the prediction model analyzes the multiple groups of feature information based on the target mode to obtain multiple first information; wherein the first information is used to predict the performance of the target customer in different dimensions; and determining the credit risk of the target customer based on the multiple first information.

[0006] According to an embodiment of the present disclosure, multiple groups of feature information are obtained based on the customer type of the target customer, including: obtaining multiple original information related to the target customer; wherein the multiple original information includes internal customer information and external information related to the target customer; classifying the multiple original information based on the customer type to obtain multiple original information groups; wherein each original information group includes at least one item of the internal customer information and the external customer information; calculating each original information group based on a preset calculation method to obtain feature information corresponding to the original information group; wherein different original information groups correspond to different calculation methods.

[0007] According to an embodiment of the present disclosure, each original information group is calculated based on a preset calculation method to obtain characteristic information corresponding to the original information group, including at least one of the following: performing nonlinear combination on the original information in the original information group to obtain the characteristic information corresponding to the original information group; performing fuzzy logic processing on the original information in the original information group to obtain the characteristic information corresponding to the original information group; performing sliding time window calculation on the original information in the original data group to obtain the characteristic information corresponding to the original information group; analyzing the internal information based on the external information to obtain the characteristic information corresponding to the original information group.

[0008] According to an embodiment of the present disclosure, the prediction model includes a first network layer and a second network layer, the number of hidden layers in the first network layer and the second network layer are different, and the first network layer and the second network layer are used to learn different feature information.

[0009] According to an embodiment of the present disclosure, a target mode of a prediction model is selected based on the customer type of a target customer, including: dividing the hidden layer in the prediction model based on the customer type to obtain a first network layer and a second network layer; wherein the feature information input into the first network layer and the second network layer is different; and / or adjusting the parameters of the neurons of the first network layer and the second network layer respectively based on the customer type.

[0010] According to an embodiment of the present disclosure, multiple groups of feature information are input into a prediction model, and the prediction model analyzes the multiple groups of feature information based on a target pattern to obtain multiple first information, including: learning the input feature information based on the target pattern to determine feature vectors for input into different network layers; wherein the feature vectors input into different network layers are different; and each network layer learns the input feature vector separately to obtain the first information of the target customer.

[0011] According to an embodiment of the present disclosure, determining the credit risk of a target customer based on multiple first information includes: determining a credit risk calculation rule according to the customer type; and calculating the multiple first information based on the credit risk calculation rule to obtain the credit risk of the target customer.

[0012] According to an embodiment of the present disclosure, the method further includes: obtaining authorization from the target customer for collecting original information; and after obtaining authorization from the target customer for collecting original information, collecting the original information of the target customer.

[0013] A second aspect of the present disclosure provides a customer credit risk prediction device, including: an acquisition module, used to obtain multiple groups of feature information based on the customer type of the target customer; wherein the multiple groups of feature information are obtained based on the original information of the target customer, and the original information corresponding to different customer types is different; a selection module, used to select a target mode of the prediction model based on the customer type and the feature information; wherein the prediction model is a multi-mode model, and the parameters of the prediction model are different under different modes; an analysis module, used to input the multiple groups of feature information into the prediction model, and the prediction model analyzes the multiple groups of feature information based on the target mode to obtain multiple first information; wherein the first information is used to predict the performance of the target customer in different dimensions; and a determination module, used to determine the credit risk of the target customer based on the multiple first information.

[0014] A third aspect of the present disclosure provides an electronic device, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0015] The fourth aspect of the present disclosure further provides a computer-readable storage medium having a computer program or instructions stored thereon, which implements the steps of the above method when the above computer program or instructions are executed by a processor.

[0016] The fifth aspect of the present disclosure further provides a computer program product, including a computer program or instructions, which implement the steps of the above method when the above computer program or instructions are executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above contents and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0018] Figure 1 A diagram schematically illustrates an application scenario of a method, apparatus, device, medium, and program product for predicting customer credit risk according to an embodiment of the present disclosure;

[0019] Figure 2 A flowchart of a method for predicting customer credit risk according to an embodiment of the present disclosure is schematically shown;

[0020] Figure 3 Schematically shows a flow chart of obtaining multiple groups of feature information based on customer types of target customers according to an embodiment of the present disclosure;

[0021] Figure 4 A flowchart schematically shows a target mode of selecting a prediction model based on the customer type of a target customer according to an embodiment of the present disclosure;

[0022] Figure 5A flowchart schematically shows a method of analyzing multiple sets of feature information based on a target mode by a prediction model to obtain multiple first information according to an embodiment of the present disclosure;

[0023] Figure 6 A flowchart of determining the credit risk of a target customer based on multiple first information according to an embodiment of the present disclosure is schematically shown;

[0024] Figure 7 A schematic diagram showing a structural block diagram of a customer credit risk prediction device according to an embodiment of the present disclosure; and

[0025] Figure 8 A block diagram of an electronic device suitable for implementing a method for predicting customer credit risk according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0026] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0027] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0028] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0029] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0030] In the technical solution of the present disclosure, the user information (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0031] In the scenario of using personal information for automated decision-making, the methods, devices, and systems provided by the embodiments of the present disclosure provide users with corresponding operation portals for users to choose to agree or reject the automated decision-making results; if the user chooses to reject, the expert decision-making process will be entered. The expression "automated decision-making" here refers to the activity of automatically analyzing and evaluating a person's behavioral habits, interests and hobbies, or economic, health, credit status, etc. through computer programs, and making decisions. The expression "expert decision-making" here refers to the activity of making decisions by people who specialize in a certain field, have specialized experience, knowledge and skills, and have reached a certain level of professionalism.

[0032] An embodiment of the present disclosure provides a method for predicting customer credit risk, including: obtaining multiple groups of feature information based on customer types of target customers; wherein the multiple groups of feature information are obtained based on original information of the target customers, and different original information corresponds to different customer types; selecting a target mode of a prediction model based on customer type and feature information; wherein the prediction model is a multi-mode model, and parameters of the prediction model are different in different modes; inputting the multiple groups of feature information into the prediction model, and the prediction model analyzes the multiple groups of feature information based on the target mode to obtain multiple first information; wherein the first information is used to predict the performance of the target user in different dimensions; and determining the credit risk of the target customer based on the multiple first information.

[0033] Figure 1 The application scenario diagram of the customer credit risk prediction method, device, equipment, medium and program product according to the embodiments of the present disclosure is schematically shown.

[0034] like Figure 1 As shown, the application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used to provide a medium for a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0035] The user can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only for example).

[0036] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0037] The server 105 may be a server that provides various services, such as a background management server (only as an example) that provides support for websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0038] It should be noted that the customer credit risk prediction method provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the customer credit risk prediction device provided in the embodiment of the present disclosure can generally be set in the server 105. The customer credit risk prediction method provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Accordingly, the customer credit risk prediction device provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105.

[0039] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0040] The following will be based on Figure 1 The scene described by Figure 2~Figure 6 The customer credit risk prediction method of the disclosed embodiment is described in detail.

[0041] Figure 2 The flowchart of the customer credit risk prediction method according to the embodiment of the present disclosure is schematically shown.

[0042] like Figure 2 As shown, the customer credit risk prediction of this embodiment includes operations S210 to S240.

[0043] In operation S210, multiple sets of feature information are obtained based on customer types of target customers; wherein the multiple sets of feature information are obtained based on original information of the target customers, and different original information corresponds to different customer types.

[0044] In some embodiments, customer types may include, for example, individual customers and corporate customers, etc. Among them, individual customers may be further subdivided into ordinary depositors, high-end investors, small and medium-sized enterprise owners, etc., and corporate customers may be further subdivided into start-ups, multinational companies, large enterprises, small and medium-sized enterprises, etc. Corresponding original information is collected for different customer types.

[0045] And, based on the preset calculation method, the original information is calculated separately to integrate the original information into multiple sets of feature information that can reflect the characteristics of target customers in different dimensions. By preprocessing the original information, the expressiveness of the data can be effectively improved, and the target customers can be described from different angles, providing a richer perspective, which helps the subsequent model to understand the data more deeply.

[0046] In the specific implementation process, the calculation methods for different original information are the same or different, and can be flexibly selected according to the type of original information, customer type and actual application scenario to improve the data fusion effect and enhance the interpretability of feature information.

[0047] In operation S220, a target mode of a prediction model is selected based on the customer type and feature information; wherein the prediction model is a multi-mode model, and parameters of the prediction model are different in different modes.

[0048] In some embodiments, the prediction model in the present disclosure is a multi-mode model, wherein the parameters and configurations of the model in each mode are different to adapt to the prediction needs of different target customers. By selecting the target mode of the prediction model, the adaptability of the prediction model can be effectively improved, the characteristic information of the target customers can be fully utilized, and the adaptability to the prediction of different customer needs can be met, thereby effectively improving the accuracy of the prediction results. And by constructing different modes for the prediction model, it is possible to effectively avoid using an overly complex model to process all types of data, effectively reducing the complexity and computational cost of the model.

[0049] In operation S230, multiple sets of feature information are input into a prediction model, and the prediction model analyzes the multiple sets of feature information based on a target mode to obtain multiple first information; wherein the first information is used to predict the performance of the target user in different dimensions.

[0050] In some embodiments, multiple sets of feature information are first preprocessed, and the preprocessed feature information is input into the prediction model, which analyzes the input multiple sets of feature information based on the target model, wherein the analysis process may include, for example, feature extraction, feature fusion, and feature selection steps to extract key information related to the credit risk prediction task, and conduct in-depth learning and analysis of the feature information through the internal mechanism of the model (such as attention mechanism, convolutional neural network, etc.) to capture the user's behavior patterns and potential needs in different dimensions. Based on the analysis results of the model, multiple first information are output to predict the performance of target customers in different dimensions, such as customer income stability, repayment ability, etc.

[0051] In operation S240, the credit risk of the target customer is determined based on the plurality of first information.

[0052] In some embodiments, multiple first information are integrated to determine the credit risk of target customers. Different target customers may have different ways of determining credit risk, which effectively improves the flexibility of credit risk determination, meets the changes in different market environments and customer needs, and improves the pertinence and effectiveness of risk management.

[0053] The customer credit risk prediction method provided by the present disclosure obtains multiple sets of feature information according to the customer type of the target customer, realizes differentiated feature information acquisition, and thus more accurately extracts features closely related to the credit status of the target customer. And based on a multi-mode prediction model, customer credit risk prediction is performed to flexibly adapt to the risk prediction needs in different scenarios, so that the model can be adaptively adjusted according to the characteristics of different customer types and feature information, so as to more accurately predict the performance of the target customer in different dimensions, thereby effectively improving the accuracy of credit risk prediction.

[0054] It should be noted that, in the embodiments of the present disclosure, before obtaining the original information of the target customer, the customer's consent or authorization may be obtained. For example, before performing operation S210, a request to obtain the original information may be issued to the target customer. When the customer agrees or authorizes that the original information may be obtained, operation S210 is performed.

[0055] Figure 3 The flowchart of obtaining multiple groups of feature information based on the customer type of the target customer according to an embodiment of the present disclosure is schematically shown.

[0056] like Figure 3 As shown, the method of obtaining multiple groups of feature information based on the customer types of target customers in this embodiment includes operations S310 to S330.

[0057] In operation S310, a plurality of original information related to a target customer is obtained; wherein the plurality of original information includes internal customer information and external information related to the target customer.

[0058] In some embodiments, the original information collected may be different for different types of customers. And the collection methods of different original information may also be different. For example, internal customer information can be obtained through internal databases, designated systems (such as customer relationship management systems), and direct customer information, and external information can be obtained through market research, third-party data providers, online public platforms, etc.

[0059] Taking individual customers as an example, the collected internal customer information may include age, occupation, education level, credit score, credit record, loan repayment history, credit card usage, consumption habits, consumption frequency, consumption amount, investment status, income status, asset status, etc. In addition, external information related to customers is collected based on internal customer information. For example, market information related to customer assets can be obtained based on the customer's asset status. Assuming that the customer's assets include real estate, cars, etc., price information and sales information similar to the real estate and cars held by the customer in the current market can be obtained.

[0060] Taking corporate customers as an example, the collected internal customer information may include the company's registered capital, establishment time, business scope, operating income, net profit, asset-liability ratio, historical performance record, industry, etc. In addition, external information related to the customer is collected based on the customer's internal information. For example, the development trend of the industry can be obtained based on the company's business scope and industry.

[0061] In operation S320, the plurality of original information are classified based on the customer type to obtain a plurality of original information groups; wherein each original information group includes at least one of the customer's internal information and external information.

[0062] In some embodiments, the multiple original information are classified based on the customer type to form multiple original information groups, wherein the original information groups may include, for example, a first original information group for reflecting the credit status of the target customer, a second original information group for reflecting the repayment ability of the target customer, and a third original information group for reflecting the asset value of the target customer.

[0063] In operation S330, calculations are performed on each original information group based on a preset calculation method to obtain characteristic information corresponding to the original information group; different original information groups correspond to different calculation methods.

[0064] In some embodiments, different calculation methods are set for different types of original information groups. These calculation methods may include statistical analysis, data mining, machine learning and other algorithms. Each original information group is calculated by the corresponding calculation method to obtain characteristic information corresponding to each original information group. These characteristic information are the refinement and summary of the original information, which can more intuitively reflect certain attributes or behavior patterns of the customer.

[0065] The embodiments of the present disclosure generate personalized feature information by integrating and classifying the original information, which helps to accurately reflect the customer's credit status, income, asset value, etc., thereby effectively improving the accuracy of subsequent credit risk prediction. When collecting original information, the present disclosure collects information from both internal and external dimensions, fully considering the impact of market changes on customer credit risk prediction, and improving the real-time and flexibility of customer credit risk prediction.

[0066] In some embodiments, in operation S330, each original information group is calculated based on a preset calculation method to obtain characteristic information corresponding to the original information group, including at least one of the following:

[0067] The original information in the original information group is nonlinearly combined to obtain the characteristic information corresponding to the original information group.

[0068] The original information in the original information group is processed by fuzzy logic to obtain the characteristic information corresponding to the original information group.

[0069] The original information in the original data group is subjected to sliding time window calculation to obtain the characteristic information corresponding to the original information group.

[0070] The internal information is analyzed based on the external information in the original data group to obtain the characteristic information corresponding to the original information group.

[0071] The operation S330 is described in detail below through specific embodiments.

[0072] Taking the original information group including credit record, loan repayment history, and credit card usage as an example, the original information in the original information group is nonlinearly combined to obtain the characteristic information corresponding to the original information group.

[0073] Among them, credit records may include, for example, the customer's loan status, credit card status, public records (such as bankruptcy records), etc.; loan repayment history may include, for example, the customer's monthly repayment record, overdue status, etc.; credit card usage status may include, for example, the customer's number of credit cards, total credit limit, used limit, repayment status, etc.

[0074] In the specific implementation process, the original information is reduced in dimension by principal component analysis, the main information is retained, and then the main information is combined by nonlinear combination module to obtain the first characteristic information used to reflect the credit status of the target customer. The weight of each original information can be dynamically adjusted based on the specific situation of the target customer.

[0075] For example, a neural network can be constructed, the input layer of which is the original information and the dynamic weights of each original information, and the nonlinear relationship between each original information is captured through the hidden layer of the neural network to output the first feature information.

[0076] The first feature information S 信用状况 The calculation formula can be expressed as:

[0077] S 信用状况 =f(W1*g1(S 信用记录 )+W2*g2(S 信用卡使用情况 )+W3*g3(S 贷款偿还历史 ))

[0078] Among them, S 信用记录 , S 信用卡使用情况 , S 贷款偿还历史 They respectively represent the scores of credit record, credit card usage, and loan repayment history, W1, W2, and W3 are the dynamic weights corresponding to credit record, credit card usage, and loan repayment history, f is the neural network model, and g1, g2, and g3 represent the nonlinear combination process of credit record, credit card usage, and loan repayment history.

[0079] In one embodiment of the present disclosure, the nonlinear combination of credit records, credit card usage, and loan repayment history includes:

[0080]

[0081] Taking the original information group containing the target customer's payroll, tax bill, and bank statement as an example, the original information in the original information group is processed through fuzzy logic and deep learning to obtain the feature information corresponding to the original information group.

[0082] In the specific implementation process, the stability of payroll, tax bill and bank statement is converted into fuzzy set, and mapped using fuzzy membership function, and the fuzzy membership function of payroll stability μ(WSS), tax bill stability fuzzy membership function μ(TSS), bank statement stability fuzzy membership function μ(BSS) are obtained. The fuzzy historical function is fuzzy reasoned using the fuzzy rule base to obtain the fuzzy reasoning result f(WSS, TSS, BSS). The fuzzy reasoning result f(WSS, TSS, BSS)=max(min(μ(WSS)),min(μ(TSS)),min(μ(BSS))), where max( ) and min( ) are the maximum value function and the minimum value function, respectively, which are used to obtain the maximum value and minimum value from multiple data. The original information and the fuzzy reasoning results are input into the deep learning model, and the deep learning model performs feature extraction and nonlinear combination on the input to obtain the second feature information CRS used to characterize the repayment ability of the target customer. The calculation formula of CRS is CRS=D(f(WSS, BSS, TSS)+α*M(WSS, BSS, TSS)), where α is a weight parameter used to balance the influence of fuzzy reasoning and deep learning models. D is a defuzzification function used to convert fuzzy results into specific values. The calculation method of the fuzzy membership function of different original information is different. Those skilled in the art can set it according to the actual situation, and this disclosure does not limit it here.

[0083] Taking the original information group containing the target customer's liability information and asset information as an example, the original information in the original information group is calculated based on the sliding event window to obtain the third characteristic information used to reflect the customer's repayment ability.

[0084] In the specific implementation process, the size of the sliding window is set according to the evaluation requirements, and the selection of the window size should be able to capture the changing trends of user load and assets. For each sliding window, the total load and asset combination in the window are calculated, and the net assets in each sliding window are calculated according to the net asset calculation formula (net assets = total assets - total liabilities). Combined with the economic scenarios corresponding to each time window, the changing trends of net assets under different economic scenarios are calculated, and the customer's net assets are evaluated based on the changing trends of net assets to obtain the third feature information used to reflect the customer's repayment ability.

[0085] Taking the asset information and market information held by the target customer in the original information group as an example, the internal information (asset information) is analyzed based on the external information (market information) to obtain the fourth characteristic information used to reflect the market value of the customer's assets.

[0086] In the specific implementation process, taking the user's asset as a real estate, market information related to the real estate is obtained, and the market value of the asset is evaluated based on the market information to obtain the fourth characteristic information used to reflect the value of the customer's asset. Specifically, the market information may include, for example, the transaction heat and price distribution of real estate in different regions, and other real estates with similar locations, years, and states to the real estate held by the user. By analyzing the transaction heat and price distribution of real estate in different regions, the relevant factors affecting the market value of the customer's assets are analyzed, and other real estates similar to the real estate held by the customer are analyzed based on the relevant factors to obtain the fourth characteristic information used to reflect the market value of the customer's assets.

[0087] The customer credit risk prediction method provided by the embodiment of the present disclosure calculates the original information group based on a specific calculation method, thereby obtaining the feature information corresponding to each original information group, which can improve the accuracy and effectiveness of feature information extraction, obtain richer feature information, and meet different feature information acquisition needs.

[0088] In some embodiments, the prediction model provided by the embodiments of the present disclosure is a multi-modal model, which includes a first network layer and a second network layer, wherein the first network layer and the second network layer are used to process different types of input features, and the number of hidden layers included in the first network layer and the second network layer is different. The depth of the first network layer is relatively shallow, and is used to process input features with relatively simple data features and a large amount of information; the depth of the second network layer is relatively deep, and is used to process input features with relatively complex data features.

[0089] The prediction model provided by the embodiment of the present disclosure is provided with multiple network layers for learning different input features respectively, which helps the model focus on learning different types of feature information and improves the flexibility of data processing of the prediction model. By setting different network models, it is possible to effectively improve resource utilization and reduce the amount of calculation and unnecessary consumption of resources.

[0090] Figure 4 A flowchart of selecting a target mode of a prediction model based on the customer type of a target customer according to an embodiment of the present disclosure is schematically shown.

[0091] like Figure 4 As shown, the target mode of selecting a prediction model based on the customer type of a target customer in this embodiment includes operations S410 to S420.

[0092] In operation S410, the hidden layer in the prediction model is divided based on the customer type to obtain a first network layer and a second network layer; wherein the feature information input into the first network layer and the second network layer is different.

[0093] In operation S420, parameters of neurons in the first network layer and the second network layer are adjusted based on the client type.

[0094] In some embodiments, the hidden layers included in the first network layer and the second network layer in the prediction model in different modes are different, and the types of input features learned by the first network layer and the second network layer and the parameters of the neurons may also be different.

[0095] In the specific implementation process, an appropriate hidden layer is selected as a dividing point based on the customer type, and the model is divided into a first network layer and a second network layer, wherein the first network layer and the second network layer learn different feature information. By dividing the hidden layer, the model can better adapt to different types of customer data and improve the generalization ability of the model. And by dividing the hidden layer into different network layers, it can effectively help the model focus on learning different types of feature information, improve the model's learning accuracy for different feature information of specific customer types, and improve the accuracy of customer credit risk prediction.

[0096] Since the key features corresponding to different customer types may be different, for example, for high-risk customers, the model may pay more attention to their default records and potential risk points, and for low-risk customers, it may consider their repayment ability and credit records more. Therefore, in addition to the different hidden layer divisions, the parameters of neurons in each network layer may also be different in the prediction models under different modes. By adjusting the parameters of neurons in the network layer, the accuracy of the prediction model's attention to key features in this mode can be improved, thereby enhancing the expressive power of features and improving the prediction accuracy of the model. Specifically, after determining the first network layer and the second network layer, the parameters of the neurons in the first network layer and the second network layer (such as weights, biases, etc.) are adjusted for different types of customers, so that the network learns different features of the input data.

[0097] The disclosed embodiment adjusts the prediction mode of the prediction model based on the customer type to achieve accurate credit risk prediction for different customers based on different modes. The network layer and neuron parameters in the prediction model are customized for different customer types, so that the model can more accurately and deeply mine and express the key characteristics of each customer, thereby providing richer information for subsequent decision-making and improving the accuracy and reliability of customer credit risk prediction. The model is flexibly adjusted according to actual needs to meet the needs of different customer types and different business scenarios, thereby improving the generalization ability and comprehensiveness of the model in practical applications.

[0098] In some embodiments, the model can be obtained by the following training method: collect and organize corresponding data sets according to customer types, wherein these data sets can reflect the characteristics and risk points of each type of customer. Input the data sets of different customer types into the model respectively, and adjust and optimize the parameters of the model during the model training process (such as adjusting the division of the hidden layer, adjusting the parameters of the neurons in the network layer), until the model reaches a stable or optimal state on the verification set of the customer type, and determine the model under the current parameters as the target mode corresponding to the customer type.

[0099] In a specific implementation process, when training the prediction model, the parameters of the neurons in the first network layer and the second network layer can be determined based on different standards.

[0100] Since the data features of the first network layer are relatively simple and contain a large amount of information, the improved entropy or information gain ratio can be used as a criterion for adjusting neuron parameters so that the first network layer learns the corresponding feature representation.

[0101] In the embodiment of the present disclosure, the calculation formula of entropy is:

[0102]

[0103] Among them, H(X) represents the entropy of the data set X, n represents the number of possible values ​​in the data set, and P(x i ) represents the probability of the occurrence of the i-th value in the data set, and b is the base. The value of b can be adjusted according to actual needs, for example, it can be e or 2.

[0104] By introducing an adjustable base b for entropy calculation, it can more flexibly adapt to different data distributions and improve the accuracy of feature selection.

[0105] In the embodiment of the present disclosure, the information gain ratio is calculated as follows:

[0106]

[0107] Among them, IG(X,Y) represents the ratio of the information gain of feature Y to data set X, H(X) is the entropy of data set X, H(X|Y) represents the conditional entropy of data set X under the condition of given feature Y, and Gini(X) represents the Gini index of data set X. The traditional information gain ratio only considers entropy, which may not work well when processing uneven data. The present disclosure can better measure the effectiveness of feature selection by combining the Gini index to calculate the information gain ratio, especially when the data is unevenly distributed.

[0108] Calculate the entropy or information gain ratio of different types of customer input features respectively, determine the input features that need to be learned in different modes, and modify the parameters of the neurons in the first network layer based on the determined input features, and set the neuron parameters to appropriate initial values ​​to improve the efficiency and accuracy of input feature learning.

[0109] As the network depth of the second network layer increases, the data features become more complex, requiring more complex calculations and faster processing speeds. Therefore, the improved Gini coefficient can be used as a criterion for adjusting neuron parameters so that the second network layer can learn the corresponding feature representations.

[0110] In the embodiment of the present disclosure, the calculation formula of the Gini index is:

[0111]

[0112] Where Gini(D) represents the purity of the dataset D, and m represents the number of categories in the dataset. i represents the weight of the i-th class, P i Represents the probability of the i-th category in the data set. The traditional Gini index treats all features equally and may ignore the differences in the importance of features. The embodiment of the present disclosure can better highlight important features and improve the accuracy of classification decisions by introducing weights.

[0113] The Gini coefficients of different types of customer input features are calculated respectively to determine the input features that need to be learned in different modes. Based on the determined input features, the parameters of the neurons in the second network layer are modified and the neuron parameters are set to appropriate initial values ​​to improve the efficiency and accuracy of input feature learning.

[0114] The disclosed embodiment uses different standards to adjust the neuron parameters of the first network layer and the second network layer, so that the first network layer and the second network layer can better adapt to different decision-making requirements, thereby improving the performance and stability of the model.

[0115] Figure 5 The flowchart schematically shows a process of analyzing multiple groups of feature information based on a target pattern by a prediction model to obtain multiple first information according to an embodiment of the present disclosure.

[0116] like Figure 5 As shown, in this embodiment, the prediction model analyzes multiple groups of feature information based on the target mode to obtain multiple first information, including operation S510 to operation S520.

[0117] In operation S510, input feature information is learned based on a target pattern to determine feature vectors for input to different network layers; wherein the feature vectors input to different network layers are different.

[0118] In operation S520, each network layer learns the input feature vector respectively to obtain the first information of the target customer.

[0119] In some embodiments, the first network layer and the second network layer may learn the input feature information, or the special layer may send the feature vector to the first network layer and the second network layer respectively.

[0120] When the feature information is input into the prediction model, the input layer of the prediction model integrates the input feature information. The first network layer and the second network layer learn the feature information in the input layer respectively to obtain the feature vector corresponding to the network layer. Specifically, the network decides how to combine and transform these input features by learning weights and biases. Different hidden layer neurons may respond to different parts of the input features, depending on their parameter settings in the target mode. By conditioning the parameters of the prediction model, different network layers focus on processing specific features or feature combinations to complete specific tasks. Alternatively, a special layer is set in the prediction model, which is used to identify and extract local features in the input layer, and then pass the extracted local features to the corresponding network layer for further processing by the network layer.

[0121] The first network layer and the second network layer learn the input feature vectors respectively to predict the performance of the target user in the current dimension, for example, the target user's income, asset changes, etc.

[0122] The disclosed embodiment analyzes the input feature information through different network layers respectively, which can effectively utilize the diversity of feature information, predict credit risk from different dimensions, and improve the accuracy and robustness of the prediction model. Setting different network layers can realize parallel processing of input information, improve the computational efficiency of the model, and shorten the response time of the model while maintaining the model performance. Learning the input feature information based on multiple network layers can effectively reduce the deviation and uncertainty that may be caused by a single network layer, and achieve a more comprehensive comprehensive decision.

[0123] Figure 6 The flowchart of determining the credit risk of a target customer based on multiple first information according to an embodiment of the present disclosure is schematically shown.

[0124] like Figure 6 As shown, the method of determining the credit risk of a target customer based on a plurality of first information in this embodiment includes operations S610 to S620.

[0125] In operation S610, a credit risk calculation rule is determined according to the customer type.

[0126] In operation S620, the plurality of first information is calculated based on a credit risk calculation rule to obtain the credit risk of the target customer.

[0127] In some embodiments, since different types of customers have different characteristics and performances in terms of credit risk, it is necessary to formulate differentiated credit risk calculation rules, set different credit risk calculation parameters and weights for different types of customers, calculate multiple first information through corresponding credit risk calculation rules, and comprehensively derive the credit risk of the target customer, so as to accurately reflect the credit risk status of different types of customers and improve the accuracy of credit risk prediction.

[0128] Based on the above customer credit risk prediction method, the present disclosure also provides a customer credit risk prediction device. Figure 7 The device is described in detail.

[0129] Figure 7 The structure block diagram of the customer credit risk prediction device according to the embodiment of the present disclosure is schematically shown.

[0130] like Figure 7 As shown, the customer credit risk prediction device 700 of this embodiment includes an acquisition module 710, a selection module 720, an analysis module 730 and a determination module 740.

[0131] The acquisition module 710 is used to obtain multiple sets of feature information based on the customer type of the target customer; wherein the multiple sets of feature information are obtained based on the original information of the target customer, and the original information corresponding to different customer types is different. In one embodiment, the acquisition module 710 can be used to perform the operation S210 described above, which will not be repeated here.

[0132] The selection module 720 is used to select a target mode of the prediction model based on the customer type and feature information; wherein the prediction model is a multi-mode model, and the parameters of the prediction model are different in different modes. In one embodiment, the selection module 720 can be used to perform the operation S220 described above, which will not be repeated here.

[0133] The analysis module 730 is used to input multiple sets of feature information into the prediction model, and the prediction model analyzes the multiple sets of feature information based on the target mode to obtain multiple first information; wherein the first information is used to predict the performance of the target customer in different dimensions. In one embodiment, the analysis module 730 can be used to perform the operation S230 described above, which will not be repeated here.

[0134] The determination module 740 is used to determine the credit risk of the target customer based on the plurality of first information. In one embodiment, the determination module 740 can be used to perform the operation S240 described above, which will not be described in detail here.

[0135] According to an embodiment of the present disclosure, any multiple modules among the acquisition module 710, the selection module 720, the analysis module 730 and the determination module 740 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the acquisition module 710, the selection module 720, the analysis module 730 and the determination module 740 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation modes of software, hardware and firmware or in a proper combination of any of them. Alternatively, at least one of the acquisition module 710, the selection module 720, the analysis module 730 and the determination module 740 can be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding function can be performed.

[0136] Figure 8 A block diagram of an electronic device suitable for implementing a method for predicting customer credit risk according to an embodiment of the present disclosure is schematically shown.

[0137] like Figure 8 As shown, the electronic device 800 according to an embodiment of the present disclosure includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage part 808 to a random access memory (RAM) 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a dedicated microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include an onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0138] In RAM 803, various programs and data required for the operation of electronic device 800 are stored. Processor 801, ROM 802 and RAM 803 are connected to each other via bus 804. Processor 801 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 802 and / or RAM 803. It should be noted that the program can also be stored in one or more memories other than ROM 802 and RAM 803. Processor 801 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in the one or more memories.

[0139] According to an embodiment of the present disclosure, the electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to the bus 804. The electronic device 800 may further include one or more of the following components connected to the input / output (I / O) interface 805: an input portion 806 including a keyboard, a mouse, etc.; an output portion 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 808 including a hard disk, etc.; and a communication portion 809 including a network interface card such as a LAN card, a modem, etc. The communication portion 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output (I / O) interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed, so that the computer program read therefrom is installed into the storage portion 808 as needed.

[0140] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0141] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, it may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus or a device. For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the ROM 802 and / or RAM 803 described above and / or one or more memories other than ROM 802 and RAM 803.

[0142] The embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the customer credit risk prediction method provided by the embodiment of the present disclosure.

[0143] The above functions defined in the system / device of the embodiment of the present disclosure are executed when the computer program is executed by the processor 801. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0144] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 809, and / or installed from a removable medium 811. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0145] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the processor 801, the above functions defined in the system of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0146] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0147] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0148] It will be appreciated by those skilled in the art that the features described in the various embodiments of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations or combinations are not explicitly described in the present disclosure. In particular, without departing from the spirit and teachings of the present disclosure, the features described in the various embodiments of the present disclosure may be combined and / or combined in a variety of ways. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0149] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A customer credit risk prediction method, characterized in that: The method comprises: Obtaining multiple sets of feature information based on the customer types of the target customers; wherein the multiple sets of feature information are obtained based on the original information of the target customers, and the original information corresponding to different customer types is different; Selecting a target mode of a prediction model based on the customer type and the feature information; wherein the prediction model is a multi-mode model, and the parameters of the prediction model are different in different modes; Input the multiple sets of feature information into the prediction model, and the prediction model analyzes the multiple sets of feature information based on the target mode to obtain multiple first information; wherein the first information is used to predict the performance of the target customer in different dimensions; The credit risk of the target customer is determined based on the plurality of first information.

2. The customer credit risk prediction method according to claim 1, characterized in that: The multiple sets of feature information are obtained based on the customer type of the target customer, including: Acquire multiple original information related to the target customer; wherein the multiple original information includes internal customer information and external information related to the target customer; Classifying the plurality of original information based on the customer type to obtain a plurality of original information groups; wherein each original information group includes at least one of the customer internal information and the external information; Each original information group is calculated based on a preset calculation method to obtain characteristic information corresponding to the original information group; wherein different original information groups correspond to different calculation methods.

3. The customer credit risk prediction method according to claim 2, characterized in that: The calculating of each original information group based on the preset calculating method to obtain the characteristic information corresponding to the original information group includes at least one of the following: Performing nonlinear combination on the original information in the original information group to obtain characteristic information corresponding to the original information group; Performing fuzzy logic processing on the original information in the original information group to obtain feature information corresponding to the original information group; Performing sliding time window calculation on the original information in the original data group to obtain feature information corresponding to the original information group; The internal information is analyzed based on the external information to obtain characteristic information corresponding to the original information group.

4. The customer credit risk prediction method according to claim 1, characterized in that: The prediction model includes a first network layer and a second network layer, the first network layer and the second network layer have different numbers of hidden layers, and the first network layer and the second network layer are used to learn different feature information.

5. The customer credit risk prediction method according to claim 4, characterized in that: The selecting the target mode of the prediction model based on the customer type of the target customer includes: Dividing the hidden layer in the prediction model based on the customer type to obtain a first network layer and a second network layer; wherein the feature information input into the first network layer and the second network layer is different; and / or The parameters of the neurons of the first network layer and the second network layer are adjusted respectively based on the client type.

6. The customer credit risk prediction method according to claim 1, characterized in that: The inputting the multiple sets of feature information into the prediction model, and the prediction model analyzing the multiple sets of feature information based on the target mode to obtain multiple first information, including: Learning the input feature information based on the target pattern to determine feature vectors for input to different network layers; wherein the feature vectors input to different network layers are different; Each network layer learns the input feature vector respectively to obtain the first information of the target customer.

7. The customer credit risk prediction method according to claim 1, characterized in that: The determining the credit risk of the target customer based on the plurality of first information includes: Determine credit risk calculation rules based on the customer type; The plurality of first information are calculated based on the credit risk calculation rule to obtain the credit risk of the target customer.

8. The method according to claim 1, characterized in that The method further comprises: Obtaining authorization from the target customer for collecting original information; After obtaining authorization from the target customer to collect original information, the original information of the target customer is collected.

9. A customer credit risk prediction device, characterized in that: The device comprises: An acquisition module, used to acquire multiple sets of feature information based on the customer types of the target customers; wherein the multiple sets of feature information are acquired based on the original information of the target customers, and different original information corresponds to different customer types; A selection module, configured to select a target mode of a prediction model based on the customer type and the feature information; wherein the prediction model is a multi-mode model, and the parameters of the prediction model are different in different modes; an analysis module, configured to input the plurality of sets of feature information into the prediction model, and the prediction model analyzes the plurality of sets of feature information based on the target mode to obtain a plurality of first information; wherein the first information is used to predict the performance of the target customer in different dimensions; and A determination module is used to determine the credit risk of the target customer based on the multiple first information.

10. An electronic device, comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 8.

11. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

12. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images