Credit risk identification method and device, electronic equipment and storage medium

Abstract

The invention discloses a credit risk identification method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining a to-be-identified customer evaluation index; according to the customer evaluation index, inputting a pre-trained risk identification model, the risk identification model being obtained after training to convergence through a neural network, and the risk identification model adopting an attention mechanism; and outputting the customer credit risk according to the risk identification model. On one hand, comprehensive analysis and feature extraction of customer information are realized, and on the other hand, credit risk identification precision is improved.

Description

Technical Field

[0001] The present application relates to the technical field of credit risk management, and in particular to a credit risk identification method, device, electronic device, and storage medium. Background Art

[0002] Commercial banks are essentially risk-taking enterprises, and their main business activities, such as accepting deposits, making loans, and participating in financial market transactions, all involve different types of risk management.

[0003] Among all the business activities of banks, credit risk management occupies an extremely important position and is the top priority of bank risk management.

[0004] As the economic environment deteriorates, the possibility of credit risk exposure increases, which may cause banks to face huge losses. How to correctly evaluate and analyze credit risks becomes increasingly important. Summary of the invention

[0005] The embodiments of the present application provide a credit risk identification method, device, electronic device, and storage medium to achieve comprehensive analysis and feature extraction of customer information, while efficiently classifying rich feature information, thereby significantly improving the accuracy of credit risk identification.

[0006] The present application embodiment adopts the following technical solutions:

[0007] In a first aspect, an embodiment of the present application provides a credit risk identification method, wherein the method comprises:

[0008] Obtain the customer evaluation indicators to be identified;

[0009] According to the customer evaluation index, a pre-trained risk identification model is input, wherein the risk identification model is obtained by training a neural network until convergence and the risk identification model adopts an attention mechanism; and

[0010] According to the risk identification model, the customer credit risk is output.

[0011] In some embodiments, the risk identification model includes:

[0012] The feature extraction layer adopts a multi-layer encoder architecture and is constructed based on a multi-dimensional indicator attention mechanism;

[0013] The classification layer adopts the MLP architecture and performs classification through the normalized function classifier.

[0014] In some embodiments, the risk identification model further includes:

[0015] Obtain multi-dimensional data evaluation indicators of customers;

[0016] Determine the corresponding classification vector data added based on the customer's multi-dimensional data;

[0017] The corresponding classification vector data and the customer multi-dimensional data evaluation index are used as the final input of the feature extraction layer of the neural network.

[0018] In some embodiments, the feature extraction layer includes a multi-dimensional indicator attention mechanism operation,

[0019] Attention operations are performed between multi-dimensional customer evaluation indicators to obtain the output of the multi-dimensional indicator attention mechanism operation.

[0020] In some embodiments, the feature extraction layer includes a skip connection operation,

[0021] Perform a skip connection and sum the output of the multi-dimensional indicator attention mechanism operation with the original input of the customer multi-dimensional data evaluation indicator to obtain a first summation result;

[0022] The first summation result is normalized and then summed with the output of the multi-dimensional indicator attention mechanism operation to obtain the output of the current encoding layer;

[0023] The output of the current coding layer is used as the input of the next coding layer, and finally after operations of multiple coding layers, the output corresponding to the classification vector is finally used as the feature output of all coding layers.

[0024] In some embodiments, the method further comprises:

[0025] Send the feature outputs of all encoding layers to the MLP layer of the MLP architecture;

[0026] After multiple layers of MLP operations in the MLP architecture, the output result is finally subjected to Softmax operation.

[0027] In some embodiments, obtaining the customer credit data to be identified includes:

[0028] Obtain any one or more groups of indicators among the customer's personal information indicators, financial status indicators, credit record indicators, behavioral characteristic indicators, personal preference indicators, and loan history indicators, and use them as the customer credit data to be identified.

[0029] In a second aspect, an embodiment of the present application further provides a credit risk identification device, wherein the device comprises:

[0030] An acquisition module, used to acquire the customer evaluation index to be identified;

[0031] An input module, used to input a pre-trained risk identification model according to the customer evaluation index, wherein the risk identification model is obtained by training a neural network until convergence and the risk identification model adopts an attention mechanism; and

[0032] The output module is used to output the customer credit risk according to the risk identification model.

[0033] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a processor; and a memory arranged to store computer executable instructions, wherein the executable instructions, when executed, cause the processor to perform the above method.

[0034] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores one or more programs. When the one or more programs are executed by an electronic device including multiple application programs, the electronic device executes the above method.

[0035] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: obtaining the customer evaluation index to be identified, and inputting the customer evaluation index into a pre-trained risk identification model, and outputting the customer credit risk according to the risk identification model. Since the risk identification model is obtained by training a neural network until convergence and the risk identification model adopts an attention mechanism, more accurate and comprehensive risk prediction can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0037] Figure 1 This is a flow chart of a credit risk identification method in an embodiment of the present application;

[0038] Figure 2 A schematic diagram of the structure of a risk identification model of a credit risk identification method in an embodiment of the present application;

[0039] Figure 3 This is a schematic diagram of the structure of a credit risk identification device in an embodiment of the present application;

[0040] Figure 4 This is a schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0042] The methods of credit risk assessment of commercial banks have gone through several stages of development. Early credit risk assessment was mainly based on the subjective judgment of experts. Evaluations were made by credit assessment experts in the bank who had undergone long-term training and had rich experience, and they made the final decision.

[0043] In the credit decision-making process, too much reliance is placed on the professional knowledge and subjective judgment of credit personnel. The emergence of statistical analysis methods such as regression analysis, multivariate discriminant analysis, Logit / Probit analysis, and nearest neighbor analysis has reduced the reliance on credit experts and can integrate multi-dimensional financial indicators to assess credit risk from a quantitative perspective. This method is more objective and accurate than previous subjective judgments, but it still has some limitations and may be too dependent on specific assumptions, such as linear relationships or specific data distributions.

[0044] With the development of machine learning technology, banks have begun to use more complex machine learning and deep learning models, such as support vector machines (SVM), random forests (RF), artificial neural networks (ANN), long short-term memory networks (LSTM), etc., to conduct credit risk assessment. These models do not rely on specific data assumptions and can handle nonlinear and high-dimensional data, so they can provide more accurate risk predictions.

[0045] In the problem of credit risk identification, there are many factors that affect credit risk and their characteristics are complex. Whether it is the traditional statistical analysis-based method, the machine learning-based method or the traditional neural network-based method, it is impossible to deeply explore the correlation between the customer's multi-dimensional evaluation indicators. In addition, the traditional calculation method has factors such as the small number of parameters, which makes it difficult to extract deep abstract features from the original data, thus affecting the classification effect.

[0046] In view of the above-mentioned deficiencies, a credit risk identification method is provided in the embodiments of the present application, which adopts a new neural network algorithm for credit risk identification. The method uses a coding layer based on a multi-dimensional indicator attention mechanism for feature extraction, and finally uses MLP for feature classification. Compared with traditional statistical analysis methods and machine learning-based methods or traditional neural network-based methods, the coding layer based on a multi-dimensional indicator attention mechanism can deeply explore the potential dependencies between customers' multi-dimensional evaluation indicators. The coding layer can also be continuously stacked, thereby expanding the number of parameters of the model, causing the model to have an emergence effect, and further improving the model's perception ability. The subsequent cooperation with MLP can efficiently classify the extracted feature information, which can significantly improve the accuracy of credit risk identification.

[0047] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0048] The present application embodiment provides a credit risk identification method, such as Figure 1 As shown, a flow chart of a credit risk identification method in an embodiment of the present application is provided, and the method at least includes the following steps S110 to S130:

[0049] Step S110, obtaining the customer evaluation index to be identified.

[0050] After a customer completes a credit transaction at a bank, the customer is handed over to the corresponding post-loan account manager for post-loan management. At the same time, the customer is included in the credit risk model monitoring scope. At the end of each day, the customer's evaluation index data in multiple dimensions is read from the bank's various systems, and the read evaluation index data is sent to the trained credit risk identification neural network model based on the multi-dimensional indicator attention mechanism. After the model is calculated, the customer's risk level can be obtained. The credit risk level given by the model is then pushed to the credit system for the post-loan account manager to check whether the customer he manages has risks. The post-loan account manager can use the risk level output by the model to assist in post-loan management decisions.

[0051] It should be noted that the customer evaluation indicators are the same in the model training stage and the model verification stage. In the training stage, they are mainly used to train the risk identification model, and in the verification stage, they are mainly used to verify the output of the risk identification model. For example, when constructing the training set, many customers' personal information indicators, financial status indicators, credit record indicators, behavioral characteristic indicators, personal preference indicators, loan history indicators and other data are collected from various related systems to construct label samples and train the model. Similarly, in the actual application process, for new loan customers, data of these dimensions should also be collected from these systems to construct the input samples of the model and input the model to obtain the prediction results. In actual applications, some indicators of some customers may not have data, and default values ​​can be used to fill them.

[0052] Step S120: input a pre-trained risk identification model according to the customer evaluation index, wherein the risk identification model is obtained by training a neural network until convergence and the risk identification model adopts an attention mechanism.

[0053] By constructing a credit risk identification neural network model based on a multi-dimensional indicator attention mechanism and training it, a credit risk identification neural network model based on a multi-dimensional indicator attention mechanism is finally obtained, which is used as a risk identification model. When training the model, risk level labels are added according to the customer's loan situation, and then the labeled data is divided into training sets, test sets, and validation sets for subsequent neural network learning. During the training process, the learning rate is set to 0.001, the optimizer uses adaptive moment estimation (Adam), and the overall accuracy (OA) and Kappa coefficient (Kappa) are used to evaluate the classification results, and a credit risk identification neural network model based on a multi-dimensional indicator attention mechanism is obtained.

[0054] In addition, the risk identification model adopts an attention mechanism to further improve the perception ability of the model, thereby extracting rich feature information carried in the input data.

[0055] When designing a traditional network, only a single indicator is considered, or the multi-dimensional evaluation indicators are simply stacked, and the potential correlation between the multi-dimensional indicators cannot be explored. However, the multi-dimensional indicator attention mechanism adopted by the network model in the embodiment of the present application can explore the dependencies between the multi-dimensional evaluation indicators of the customer during the feature extraction process. In this way, the multi-dimensional evaluation indicators of the customer can be fully utilized, and a better risk identification effect can be achieved.

[0056] Step S130: outputting the customer's credit risk according to the risk identification model.

[0057] Based on the pre-trained credit risk identification network model, the risk of customers who subsequently handle credit business can be identified and the customer's credit risk can be output.

[0058] Compared with the previous credit risk identification and classification method based on the subjective judgment of experts, the risk identification model in the above method does not rely on the professional knowledge and subjective judgment of credit personnel.

[0059] Compared with statistical analysis-based methods, the risk identification model in the above method does not rely on data distribution, can cope with different credit risk classification scenarios, and is more robust.

[0060] Compared with other credit classification methods based on machine learning or deep learning, the risk identification model in the above method has stronger feature extraction capabilities and can automatically learn and extract deep abstract features through multi-layer structures. The use of attention mechanism can capture the complex dependencies between multi-dimensional input data, which can achieve more accurate and comprehensive risk prediction.

[0061] In one embodiment of the present application, the risk identification model includes: a feature extraction layer, which adopts a multi-layer encoder architecture and is constructed based on a multi-dimensional indicator attention mechanism; a classification layer, which adopts an MLP architecture and performs classification through a normalized function classifier.

[0062] like Figure 2 As shown in , it is improved on the basis of the traditional Transform to construct a credit risk identification neural network model based on the multi-dimensional indicator attention mechanism. The input of the encoding layer is the customer's multi-dimensional evaluation indicators, which are sent to the encoding layer for calculation after being standardized. The feature extraction layer is a multi-layer encoder architecture based on the multi-dimensional indicator attention mechanism. The classification layer adopts the MLP architecture and cooperates with the Softmax classifier to complete the classification. Compared with the traditional neural network model, the multi-dimensional indicator attention mechanism of the model encoding layer can capture the complex dependencies between multi-dimensional input variables. The multi-layer stacked encoding layers can expand the model parameters and further improve the perception ability of the model, thereby extracting the rich feature information contained in the input data. The classification layer can efficiently classify the rich feature information through the combination of multiple hidden layers in the MLP, which can ultimately significantly improve the accuracy of credit risk identification. As shown in Figure 2 As shown in the figure, encoder is the encoder, Self-Attention is the attention mechanism, Norm is the standard data, and Multilayer perception is the multi-perception layer.

[0063] In one embodiment of the present application, the risk identification model also includes: obtaining customer multi-dimensional data evaluation indicators; determining corresponding classification vector data added based on the customer multi-dimensional data; and using the corresponding classification vector data and the customer multi-dimensional data evaluation indicators as the final input of the feature extraction layer of the neural network.

[0064] For the customer's multi-dimensional evaluation index A = {a i,j}∈R M×N , M represents the number of customer evaluation indicator dimensions, N represents the number of customer evaluation indicators in each dimension, a i,j Represents the jth evaluation indicator of the i-th customer data dimension.

[0065] Inspired by BERT’s Class Token mechanism, we add corresponding class token data to the input customer multi-dimensional data, t = {tn}∈R N Represents the classification vector data, which is combined with the customer's multi-dimensional evaluation index as the input of the subsequent network B = {a i,j}∈R (M+1)×N , and then the network input is normalized and sent to the encoding layer for operation.

[0066] There are correlations between the evaluation indicators of different dimensions of customers. For example, there are potential correlations between the characteristics of customer personal information dimension indicators, financial information dimension indicators, customer credit dimension indicators, etc. When evaluating customer credit risk, while analyzing these indicators, it is also necessary to analyze the potential correlations between indicators of different dimensions. By adopting a multi-dimensional indicator attention mechanism, the dependency relationship between the customer's multi-dimensional evaluation indicators can be mined during the feature extraction process. Subsequently, the MLP-based classification network is used to classify the features extracted from the previous multiple coding layers, which can effectively identify the customer's credit risk. Compared with the existing credit risk identification model, the risk identification model in the embodiment of the present application can make full use of the customer's multi-dimensional evaluation indicators and can achieve better risk identification effects.

[0067] It should be noted that BERT is a pre-trained model based on the Transformer architecture. It is only used as an example and is not intended to limit the scope of protection in the embodiments of this application.

[0068] In one embodiment of the present application, the feature extraction layer includes a multi-dimensional indicator attention mechanism operation, which performs attention operations on multi-dimensional customer evaluation indicators to obtain the output of the multi-dimensional indicator attention mechanism operation.

[0069] like Figure 2 As shown in Figure 2, the core operation of the encoding layer can be divided into two parts, namely the multi-dimensional indicator attention mechanism operation and the skip connection operation. The calculation process of the multi-dimensional indicator attention mechanism is as follows:

[0070] In order to perform attention operations between the multi-dimensional evaluation indicators of the input customers, it is necessary to construct three weight matrices W with learnable parameters q , W k and W v , where W q ={w i,j}∈R N×N ,W k ={w i,j}∈R N×N ,W v ={w i,j}∈R N×N First, the network input B and the weight matrix W q Do the inner product of the matrix and get the output Q = {q i,j}∈R(M+1)×N matrix

[0071] Q= <B,W q >

[0072] Similarly, the network input B is compared with the weight matrix W k , W v Doing the inner product of the matrix, we can get the output K = {k i,j}∈R (M+1)×N and V = {v i,j}∈R (M+1)×N The matrix is ​​finally processed by the following operations to obtain the output of the multi-dimensional indicator attention mechanism operation:

[0073]

[0074] where d k Represents the dimensions of the input key vector.

[0075] In one embodiment of the present application, the feature extraction layer includes a skip connection operation, which performs a skip connection summation on the output of the multidimensional indicator attention mechanism operation and the original input of the customer's multidimensional data evaluation index to obtain a first summation result; the first summation result is standardized and then summed with the output of the multidimensional indicator attention mechanism operation to obtain the output of the current coding layer; the output of the current coding layer is used as the input of the next coding layer, and finally, after operations on multiple coding layers, the output corresponding to the classification vector is used as the feature output of all coding layers.

[0076] The coding layer jump connection operation process is as follows:

[0077] like Figure 2 As shown in the figure, after the multi-dimensional indicator attention mechanism is operated, the output of the attention mechanism is skip-connected and summed with the original input, and then the summed result is standardized and summed again with the data before the summation to obtain the output of the encoding layer. This output is used as the input of the next encoding layer. After the operation of multiple encoding layers, the output o corresponding to the classification vector aggregates the complex dependency features between the multi-dimensional input variables of the customer, and finally the output o corresponding to the classification vector is used as the feature output of all encoding layers.

[0078] In one embodiment of the present application, the method further includes: sending the feature outputs of all the coding layers to the MLP layer of the MLP architecture; after the multi-layer MLP operation in the MLP architecture, finally performing Softmax operation on the output result.

[0079] like Figure 2 As shown, the feature output o obtained by the encoding layer operation is sent to the MLP layer. Taking the MLP with two hidden layers as an example, the calculation process is as follows:

[0080] z=softmax(W (3) (σ(W (2) (σ(W (1) o+b (1) ))+b (2) ))+b (3) )

[0081] Where W (n) represents the weight matrix of the nth layer, b (n) It represents the bias term of the output of the nth layer, and the activation function is σ. After multiple layers of MLP operations, the output result is finally used as a Softmax budget, and the error back propagation algorithm is continuously iterated to obtain the final trained model. z is the output result of the forward operation of the model.

[0082] The present invention proposes a credit risk identification neural network based on a multi-dimensional indicator attention mechanism. The network includes an encoding layer based on a multi-dimensional indicator attention mechanism and a classification layer based on MLP. The multi-dimensional indicator attention mechanism of the encoding layer can capture the complex dependencies between multi-dimensional input variables. The multi-layer stacked encoding layer can expand the model parameter amount, further improve the perception ability of the model, and then extract the rich feature information contained in the input data; the classification layer can efficiently classify the rich feature information through the combination of multiple hidden layers in the MLP, and finally significantly improve the accuracy of credit risk identification.

[0083] In one embodiment of the present application, obtaining the customer credit data to be identified includes: obtaining any one or more groups of indicators among the customer's personal information indicators, financial status indicators, credit record indicators, behavioral characteristic indicators, personal preference indicators, and loan history indicators, and using them as the customer credit data to be identified.

[0084] Obtain multi-dimensional evaluation indicators of existing customers and build a sample set. The main indicators include the following:

[0085] (1) Personal information indicators: including but not limited to age, gender, marital status, education level, occupation, etc. These indicators can reflect the basic personal situation and stability of the customer.

[0086] (2) Financial status indicators: including but not limited to income, expenditure, deposits, liabilities, etc. These indicators can reflect the customer's financial status and debt repayment ability.

[0087] (3) Credit record indicators: including but not limited to credit rating, credit card usage, overdue status, etc. These indicators can reflect the customer’s credit record and credit risk.

[0088] (4) Behavioral characteristic indicators: including but not limited to consumption habits, online shopping, social media usage, etc. These indicators can reflect customers’ behavioral characteristics and lifestyles.

[0089] (5) Personal preference indicators: including but not limited to investment preferences, consumption preferences, lifestyle, etc. These indicators can reflect the customer’s personal preferences and needs.

[0090] (6) Loan history indicators: the number of loan applications in the previous year, the total amount of loans in the previous year, overdue status, etc. These indicators can reflect the customer's historical loan situation.

[0091] By dividing customer information into different dimensions and selecting representative indicators for each dimension, we can achieve comprehensive analysis and feature extraction of customer information by mining the potential dependencies between indicators of different dimensions.

[0092] Through the above method, the input data dimensions used to train the model are richer and the risk classification model is more advanced. The input data takes into account multi-dimensional evaluation indicators such as the customer's financial dimension, credit status dimension, and social and economic environment dimension. Compared with data in a single transaction dimension, multi-dimensional input data considers more comprehensive factors and will also improve the robustness of subsequent classification models.

[0093] The present application embodiment also provides a credit risk identification device 300, such as Figure 3 As shown, a schematic diagram of the structure of a credit risk identification device in an embodiment of the present application is provided. The credit risk identification device 300 at least includes: an acquisition module 310, an input module 320 and an output module 340, wherein:

[0094] In one embodiment of the present application, the acquisition module 310 is specifically used to: acquire the customer evaluation index to be identified.

[0095] After a customer completes a credit transaction at a bank, the customer is handed over to the corresponding post-loan account manager for post-loan management. At the same time, the customer is included in the credit risk model monitoring scope. At the end of each day, the customer's evaluation index data in multiple dimensions is read from the bank's various systems, and the read evaluation index data is sent to the trained credit risk identification neural network model based on the multi-dimensional indicator attention mechanism. After the model is calculated, the customer's risk level can be obtained. The credit risk level given by the model is then pushed to the credit system for the post-loan account manager to check whether the customer he manages has risks. The post-loan account manager can use the risk level output by the model to assist in post-loan management decisions.

[0096] It should be noted that the customer evaluation indicators are the same in the model training stage and the model verification stage. In the training stage, they are mainly used to train the risk identification model, and in the verification stage, they are mainly used to verify the output of the risk identification model.

[0097] In one embodiment of the present application, the input module 320 is specifically used to: input a pre-trained risk identification model according to the customer evaluation index, wherein the risk identification model is obtained by training a neural network until convergence and the risk identification model adopts an attention mechanism.

[0098] By constructing a credit risk identification neural network model based on a multi-dimensional indicator attention mechanism and training it, we finally obtain a credit risk identification neural network model based on a multi-dimensional indicator attention mechanism, which is used as a risk identification model. When training the model, we add risk level labels according to the customer's loan situation, and then divide the labeled data into training sets, test sets, and validation sets for subsequent neural network learning.

[0099] In addition, the risk identification model adopts an attention mechanism to further improve the perception ability of the model, thereby extracting rich feature information carried in the input data.

[0100] In one embodiment of the present application, the output module 340 is specifically used to: output the customer credit risk according to the risk identification model.

[0101] Based on the pre-trained credit risk identification network model, the risk of customers who subsequently handle credit business can be identified and the customer's credit risk can be output.

[0102] In one embodiment of the present application, the risk identification model includes:

[0103] The feature extraction layer adopts a multi-layer encoder architecture and is constructed based on a multi-dimensional indicator attention mechanism;

[0104] The classification layer adopts the MLP architecture and performs classification through the normalized function classifier.

[0105] In one embodiment of the present application, the risk identification model further includes:

[0106] Obtain multi-dimensional data evaluation indicators of customers;

[0107] Determine the corresponding classification vector data added based on the customer's multi-dimensional data;

[0108] The corresponding classification vector data and the customer multi-dimensional data evaluation index are used as the final input of the feature extraction layer of the neural network.

[0109] In one embodiment of the present application, the feature extraction layer includes a multi-dimensional indicator attention mechanism operation,

[0110] Attention operations are performed between multi-dimensional customer evaluation indicators to obtain the output of the multi-dimensional indicator attention mechanism operation.

[0111] In one embodiment of the present application, the feature extraction layer includes a skip connection operation,

[0112] Perform a skip connection and sum the output of the multi-dimensional indicator attention mechanism operation with the original input of the customer multi-dimensional data evaluation indicator to obtain a first summation result;

[0113] The first summation result is normalized and then summed with the output of the multi-dimensional indicator attention mechanism operation to obtain the output of the current encoding layer;

[0114] The output of the current coding layer is used as the input of the next coding layer, and finally after operations of multiple coding layers, the output corresponding to the classification vector is finally used as the feature output of all coding layers.

[0115] In one embodiment of the present application, it further includes: a processing module for

[0116] Send the feature outputs of all encoding layers to the MLP layer of the MLP architecture;

[0117] After multiple layers of MLP operations in the MLP architecture, the output result is finally subjected to Softmax operation.

[0118] In one embodiment of the present application, the acquisition module 310 is also used to

[0119] Obtain any one or more groups of indicators among the customer's personal information indicators, financial status indicators, credit record indicators, behavioral characteristic indicators, personal preference indicators, and loan history indicators, and use them as the customer credit data to be identified.

[0120] It can be understood that the above-mentioned credit risk identification device can implement each step of the credit risk identification method provided in the above-mentioned embodiment, and the relevant explanations about the credit risk identification method are applicable to the credit risk identification device, which will not be repeated here.

[0121] Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 4 At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. Of course, the electronic device may also include hardware required for other services.

[0122] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0123] The memory is used to store the program. Specifically, the program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provides instructions and data to the processor.

[0124] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a credit risk identification device at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations:

[0125] Obtain the customer evaluation indicators to be identified;

[0126] According to the customer evaluation index, a pre-trained risk identification model is input, wherein the risk identification model is obtained by training a neural network until convergence and the risk identification model adopts an attention mechanism; and

[0127] According to the risk identification model, the customer credit risk is output.

[0128] The above application Figure 1The method performed by the credit risk identification device disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by a hardware integrated logic circuit in the processor or an instruction in the form of software. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0129] The electronic device may also perform Figure 1 The method is implemented by a credit risk identification device in the Figure 1 The functions of the illustrated embodiment will not be described in detail in the embodiments of the present application.

[0130] The present application also provides a computer-readable storage medium, which stores one or more programs, wherein the one or more programs include instructions, which, when executed by an electronic device including multiple application programs, enable the electronic device to execute Figure 1 The method performed by the credit risk identification device in the illustrated embodiment is specifically used to perform:

[0131] Obtain the customer evaluation indicators to be identified;

[0132] According to the customer evaluation index, a pre-trained risk identification model is input, wherein the risk identification model is obtained by training a neural network until convergence and the risk identification model adopts an attention mechanism; and

[0133] According to the risk identification model, the customer credit risk is output.

[0134] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0135] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0136] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0137] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0138] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0139] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0140] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0141] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0142] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0143] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A credit risk identification method, wherein: The method comprises: Obtain the customer evaluation indicators to be identified; According to the customer evaluation index, a pre-trained risk identification model is input, wherein the risk identification model is obtained by training a neural network until convergence and the risk identification model adopts an attention mechanism; and According to the risk identification model, the customer credit risk is output.

2. The method of claim 1, wherein: The risk identification model includes: The feature extraction layer adopts a multi-layer encoder architecture and is constructed based on a multi-dimensional indicator attention mechanism; The classification layer adopts the MLP architecture and performs classification through the normalized function classifier.

3. The method according to claim 1 or 2, wherein: The risk identification model further includes: Obtain multi-dimensional data evaluation indicators of customers; Determine the corresponding classification vector data added based on the customer's multi-dimensional data; The corresponding classification vector data and the customer multi-dimensional data evaluation index are used as the final input of the feature extraction layer of the neural network.

4. The method of claim 2, wherein: The feature extraction layer includes a multi-dimensional indicator attention mechanism operation, Attention operations are performed between multi-dimensional customer evaluation indicators to obtain the output of the multi-dimensional indicator attention mechanism operation.

5. The method of claim 4, wherein: The feature extraction layer includes a skip connection operation, Perform a skip connection and sum the output of the multi-dimensional indicator attention mechanism operation with the original input of the customer multi-dimensional data evaluation indicator to obtain a first summation result; The first summation result is normalized and then summed with the output of the multi-dimensional indicator attention mechanism operation to obtain the output of the current encoding layer; The output of the current coding layer is used as the input of the next coding layer, and finally after operations of multiple coding layers, the output corresponding to the classification vector is finally used as the feature output of all coding layers.

6. The method according to claim 5, further comprising: Send the feature outputs of all encoding layers to the MLP layer of the MLP architecture; After multiple layers of MLP operations in the MLP architecture, the output result is finally subjected to Softmax operation.

7. The method of claim 1, wherein: The step of obtaining the customer credit data to be identified includes: Obtain any one or more groups of indicators among the customer's personal information indicators, financial status indicators, credit record indicators, behavioral characteristic indicators, personal preference indicators, and loan history indicators, and use them as the customer credit data to be identified.

8. A credit risk identification device, wherein: The device comprises: An acquisition module, used to acquire the customer evaluation index to be identified; An input module, used to input a pre-trained risk identification model according to the customer evaluation index, wherein the risk identification model is obtained by training a neural network until convergence and the risk identification model adopts an attention mechanism; and The output module is used to output the customer credit risk according to the risk identification model.

9. An electronic device, comprising: processor; as well as A memory arranged to store computer executable instructions, which when executed cause the processor to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including a plurality of application programs, causes the electronic device to execute any one of the methods of claims 1 to 7.

Images