CN115186514A

CN115186514A - Credit risk model prediction method and device

Info

Publication number: CN115186514A
Application number: CN202211032431.7A
Authority: CN
Inventors: 宋伊环
Original assignee: Bank of China Financial Technology Co Ltd
Current assignee: Bank of China Financial Technology Co Ltd
Priority date: 2022-08-26
Filing date: 2022-08-26
Publication date: 2022-10-14

Abstract

The application provides a credit risk model prediction method and a credit risk model prediction device, which are applied to the technical field of data processing and comprise the following steps: acquiring data to be processed, wherein the data to be processed comprises financial data of a customer; inputting the data to be processed into a screening model to obtain screening data; the screening model comprises a penalty function, and the penalty function is used for removing derivative data in the data to be processed; credit risk model prediction is performed using the screening data. Therefore, a punishment function is adopted to remove part of data to be processed before screening, and part of coefficients are compressed, so that the interpretability of the model can be improved, the multiple collinearity influence is reduced, the dimension of the model is reduced, overfitting is avoided, the cost of credit risk modeling prediction is reduced, and the modeling efficiency is improved.

Description

Credit risk model prediction method and device

Technical Field

The application relates to the technical field of data processing, in particular to a credit risk model prediction method.

Background

Credit services have been pursued by more and more people in recent years. For a bank, before the payment due expires, it is likely that significant adverse changes in the borrower's financial business status will affect their performance capabilities, creating a credit risk to the bank. Because of the uncertainty of the financial and business conditions of the borrower, the credit risk becomes the main risk faced by the bank, which forms the key and difficult points of the bank risk management and is the core of the bank risk management. How to predict credit risk is currently critical.

The industry proposes a method for constructing risk model prediction to predict credit risk, but the credit risk modeling characteristic factors need to be screened in the prediction process. The conventional screening method mainly adopts conventional screening logic, the workload of the screening process is large, the calculation cost is high, indexes such as sub-boxes, evidence Weight (WOE) and Information Value (IV) values of all processed characteristic factors need to be calculated and analyzed one by one, and different modeling variables need to be tried repeatedly, so that the problems of large modeling time cost and low modeling efficiency are caused.

Therefore, how to reduce the cost of credit risk modeling prediction and improve the modeling efficiency is a technical problem that needs to be solved urgently by those skilled in the art.

Disclosure of Invention

In view of this, the embodiments of the present application provide a credit risk model prediction method and apparatus, which aim to reduce modeling cost of a credit model and improve modeling efficiency.

In a first aspect, an embodiment of the present application provides a credit risk model prediction method, including:

acquiring data to be processed, wherein the data to be processed comprises financial data of a customer;

inputting the data to be processed into a screening model to obtain screening data; the screening model comprises a penalty function, and the penalty function is used for removing derivative data in the data to be processed;

credit risk model prediction is performed using the screening data.

Optionally, before the to-be-processed data is input into the screening model to obtain the screening data, the method further includes:

and preprocessing the data to be processed, wherein the preprocessing comprises repairing abnormal data, and the abnormal data comprises missing data and repeated data.

Optionally, the screening model includes: a LASSO regression model that includes the penalty function.

Optionally, the credit risk model prediction using the screening data includes:

and performing credit risk model prediction on the screening data by using a logistic regression algorithm.

Optionally, the method further includes:

obtaining results of the credit risk model prediction;

converting the result into a score, the score reflecting a condition predicted by the credit risk model;

and responding to the score value lower than a preset threshold value, and outputting an early warning prompt.

In a second aspect, an embodiment of the present application provides a credit risk model prediction apparatus, including:

the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring data to be processed, and the data to be processed comprises financial data of a client;

the screening module is used for inputting the data to be processed into a screening model to obtain screening data; the screening model comprises a penalty function, and the penalty function is used for removing part of the data to be processed;

a prediction module to use the screening data to make credit risk model predictions.

Optionally, the apparatus further comprises:

the preprocessing module is used for preprocessing the data to be processed, wherein the preprocessing comprises repairing abnormal data, and the abnormal data comprises missing data and repeated data.

Optionally, the apparatus further comprises:

a result module to obtain a result of the credit risk model prediction;

a scoring module to convert the results to a score representing a condition predicted by the credit risk model;

and the early warning module is used for responding to the condition that the score is lower than a preset threshold value and outputting early warning prompt.

In a third aspect, an embodiment provides an apparatus comprising a memory for storing instructions or code and a processor for executing the instructions or code to cause the apparatus to perform the credit risk model prediction method of any one of the preceding first aspects.

In a fourth aspect, an embodiment of the present application provides a computer storage medium, where codes are stored, and when the codes are executed, a device running the codes implements the credit risk model prediction method according to any one of the foregoing first aspects.

The embodiment of the application provides a credit risk model prediction method and a credit risk model prediction device, when the method is executed, to-be-processed data is obtained firstly, the to-be-processed data comprises financial data of a customer, and then the to-be-processed data is input into a screening model to obtain screening data; the screening model comprises a penalty function, the penalty function is used for removing derivative data in the data to be processed, and finally credit risk model prediction is carried out by using the screening data. Therefore, by adopting the penalty function to remove part of data to be processed during screening and compressing part of coefficients, the interpretability of the model can be improved, the multiple collinearity influence is reduced, the dimension of the model is reduced, overfitting is avoided, the cost of credit risk model prediction is reduced, and the modeling efficiency is improved.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments or the prior art, the drawings used in the embodiments or the prior art descriptions will be briefly described below, and obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative efforts.

FIG. 1 is a flow chart of a method of a credit risk model prediction method provided in an embodiment of the present application;

FIG. 2 is a flow chart of another method of a credit risk model prediction method provided in an embodiment of the present application;

fig. 3 is a schematic structural diagram of a credit risk model prediction apparatus according to an embodiment of the present application.

Detailed Description

The types of credit risk may be generally divided into two categories, marketable risk and non-marketable risk. The market risk mainly comes from the production and sales risk of an enterprise (borrower) (namely, the risk caused by the change of factors such as market conditions and production technologies during the production and sales of commodities by the borrower; the non-market risk mainly refers to natural and social risks.

The prevention of credit risk of commercial banks is mainly the prevention of bad credit. Bank credit assets fall into five categories: normal, concern, secondary, suspect, loss. Poor credit refers primarily to secondary, suspicious, and loss-like credit. The bank credit risk refers to the possibility of suffering asset loss due to the deviation of an actual income result and an expected income target in the operation and management process of a bank under the influence of various uncertain factors. Credit risk refers to the possibility that a borrowing enterprise will lose bank funds because it cannot return the credit information on time for various reasons. The major credit business in the bank credit business is the credit business, credit has the characteristics of higher risk and outstanding income, and the operation of the whole bank is greatly promoted. Therefore, predicting credit risk is significant.

The industry proposes a method for constructing risk model prediction to predict credit risk, but the credit risk modeling characteristic factors need to be screened in the prediction process. The conventional screening method mainly adopts conventional screening logic, the workload of the screening process is large, the calculation cost is high, indexes such as the subboxes, the Evidence Weight (WOE) and the Information Value (IV) of all processed characteristic factors need to be calculated and analyzed one by one, and different modeling variables need to be tried repeatedly, so that the problems of large modeling time cost and low modeling efficiency are caused.

The method provided by the embodiment of the application is executed by computer equipment and used for predicting the credit risk model.

It should be apparent that the described embodiments are only some of the embodiments of the present application, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application without making any creative effort belong to the protection scope of the present application.

Referring to fig. 1, fig. 1 is a flowchart of a credit risk model prediction method provided in an embodiment of the present application, including:

step S101: and acquiring data to be processed.

When the credit risk model is predicted, the characteristic factors for modeling the credit risk need to be screened, and different modeling variables need to be tried repeatedly. The acquired data to be processed, including customer financial data, is data containing all the characteristic factors of the credit risk modeling. Because the number of the characteristic factors is large, if the characteristic factors are simultaneously input into a model, the model interpretability is reduced, and the model complexity is increased, so that variable analysis is required to screen the characteristic factors.

Because the data to be processed contains thousands of characteristic factors, and the factors for screening the characteristic factors are also various, it is necessary to remove part of the characteristic factors, that is, part of the data to be processed, while screening the data to be processed.

As a possible implementation method, the client financial data includes multiple dimensions of data such as client credit, transaction, financial report, industry and government, and derivative data thereof, specifically including basic information, financial behavior, operation information, industry and government, financial information, etc. of the client, which respectively constitute different characteristic factors.

As a possible implementation method, the factors for screening the characteristic factors include: predictive power of variables, correlation between variables, simplicity of variables, robustness of variables, interpretability of variables, and the like. Among the most dominant and direct metrics are the predictive power of the variables.

Step S102: inputting the data to be processed into a screening model to obtain screening data; the screening model comprises a penalty function, and the penalty function is used for removing derivative data in the data to be processed.

Penalty functions, also known as penalty functions, are a class of restriction functions. The constraint function for constraint nonlinear programming is called penalty function, and is a class of constraint function method for converting the problem of solving constraint nonlinear programming into a series of unconstrained minimization problems. Specifically, the penalty function selects a series of incremental penalty factors, then makes a corresponding penalty function series, and finally converts the solving problem into solving a series of unconstrained minimization problems.

The screening model in the embodiment is a model comprising a penalty function, and the model can compress the coefficient of partial derivative data by using the characteristics of the penalty function, set the coefficient of partial derivative data to be zero, so as to remove the derivative data in the data to be processed, which is equivalent to screening the model by using the characteristic factors and not used as a final model entering variable.

In addition, because in the penalty function, the condition number of the matrix for the augmented objective function increases with the increase or decrease of the penalty factor, if the unconstrained problem is made to approach the original constrained problem, the penalty factor should be selected as large as possible; however, if the difficulty in solving the unconstrained problem is reduced, a smaller penalty factor should be selected, otherwise the matrix morbidity is increased. Furthermore, the number of the final mold entering variables can be adjusted according to the punishment coefficient set by the punishment function, and a relatively refined model is finally obtained.

In conclusion, the screening model is a biased estimation model capable of processing complex common data, retains the advantage of subset shrinkage, removes part of data to be processed during screening, compresses part of coefficients, can improve model interpretability, reduces multiple collinearity influence, reduces model dimensionality, and avoids overfitting.

Step S103: credit risk model prediction is performed using the screening data.

And performing credit risk modeling by using the screening data to predict credit risk. Because the screening data is data that uses a penalty function to remove part of the data to be processed during screening, the amount of data may be reduced to some extent. When the credit risk model is predicted by using the data, the time required by modeling can be reduced, the modeling cost is reduced, and the modeling efficiency is improved.

In summary, in the embodiment, the penalty function is used to remove part of data to be processed during screening, and compress part of coefficients, so that the interpretability of the model can be improved, the multiple co-linear influence can be reduced, the dimension of the model can be reduced, overfitting can be avoided, the cost of credit risk modeling prediction can be reduced, and the modeling efficiency can be improved.

In the embodiment of the present application, there are many possible implementations of the steps described in fig. 1, which are described below separately. It should be noted that the implementation manners given in the following description are only exemplary illustrations, and do not represent all implementation manners of the embodiments of the present application.

Referring to fig. 2, another method flowchart of the credit risk model prediction method provided in the embodiments of the present application is shown.

Step S201: and acquiring data to be processed.

This step is the same as S101 in the first embodiment, and therefore is not described herein again.

Step S202: and preprocessing the data to be processed.

The data to be processed is preprocessed, and abnormal data in the data to be processed is repaired, so that adverse effects of the abnormal data on a prediction result can be avoided. Wherein the abnormal data comprises missing data and repeated data. That is, in the preprocessing process, missing data needs to be supplemented, and duplicate data needs to be deduplicated. As a possible implementation manner, a prompt can be output when abnormal data occurs, so that the abnormal data can be corrected and repaired manually.

Step S203: and inputting the data to be processed into an LASSO regression model to obtain screening data.

The LASSO regression model includes the LASSO function, and the LASSO function makes some coefficients smaller, and even some coefficients with smaller absolute values become 0 directly, so it is especially suitable for the reduction of the number of parameters and the selection of parameters, and thus it is used to estimate the linear model of sparse parameters. However, the Lasso regression has a big problem, which causes that we need to carry it out alone, namely that its loss function is not continuously derivable, and the L1 norm is the sum of absolute values, which causes the loss function to have an unguided point. The method can screen variables and reduce the complexity of the model, and the variables are selectively put into the model instead of putting all the variables into the model for fitting, so that better performance parameters are obtained.

Meanwhile, complexity adjustment of the LASSO function can control the complexity of the model through a series of parameters, so that overfitting is avoided. In the process of finding the set of independent variables with the most powerful explanatory power for the dependent variables, the LASSO algorithm can perform variable screening while fitting the generalized linear model.

Screening using the LASSO regression model is one type of compressed estimation. It obtains a more refined model by constructing a penalty function, so that it compresses some regression coefficients, i.e. the sum of the absolute values of the forcing coefficients is less than a certain fixed value; while some regression coefficients are set to zero. The advantage of subset puncturing is thus retained, and is a way to process biased estimates of data with complex collinearity. In this embodiment, the coefficient of the derived data is set to 0 by using a penalty function, which may be used to remove the derived data from the data to be processed.

As a possible implementation, the LASSO regression model may be used very well in the case of linear models where the target variables are continuous, binary or multivariate discrete variables. In the risk modeling process with more derived variables, the calculation cost can be reduced by using the LASSO as a variable screening and model complexity reducing means, and more manual screening cost is avoided, so that the modeling efficiency is improved.

Step S204: and performing credit risk model prediction on the screening data by using a logistic regression algorithm.

Logistic Regression (Logistic Regression) is a Regression problem for processing a dependent variable as a classification variable, is usually a binary classification or binomial distribution problem, generally determines the probability of a classification to which a certain sample event belongs, can also process a multi-classification problem, and belongs to a classification method.

The reason why the logistic regression algorithm is used for credit risk model prediction of the screening data is that the problem of credit risk model prediction is a two-classification problem, namely that the prediction result is at risk or no risk, and the logistic regression algorithm can be used for obtaining the prediction result with the probability between 0 and 1, thereby being beneficial to the use and the explanation of the subsequent steps.

Step S205: and acquiring a result predicted by the credit risk model, and converting the result into a score.

The credit risk model predicts the probability between 0 and 1, so that other customers can further know the risk of the customers, and the prediction result can be converted into a score which can visually display the prediction result according to a certain proportion and a preset calculation formula. After the credit risk model is converted into the corresponding score, the client can clearly know the prediction result of the credit risk model, and the client can conveniently perform different measures according to different monthly measurement results.

Step S206: and responding to the score value lower than a preset threshold value, and outputting an early warning prompt.

When the score is lower than the preset threshold, the predicted risk value exceeds the bearing capacity, manual intervention is needed, therefore, early warning reminding needs to be output, and the influence of overhigh credit risk on company benefits can be prevented.

In summary, the data is preprocessed before being screened, so that the influence of abnormal data on the prediction result can be avoided, and the accuracy of the prediction result is improved; part of data to be processed is removed by adopting an LASSO regression model during screening, and part of coefficients are compressed, so that the interpretability of the model can be improved, the influence of multiple collinearity is reduced, the dimension of the model is reduced, and overfitting is avoided; the prediction result is converted into a score, so that the client can clearly know the prediction result of the credit risk model, and the client can conveniently perform different countermeasures aiming at different monthly test results; when the score is too low, the early warning prompt is output, so that the influence of too high credit risk on the benefits of the company can be prevented. The whole embodiment can reduce the cost of credit risk modeling prediction and improve the modeling efficiency.

The above provides some specific implementation manners of the credit risk model prediction method for the embodiments of the present application, and based on this, the present application also provides corresponding apparatuses. The device provided by the embodiment of the present application will be described in terms of functional modularity.

Referring to the schematic diagram of the credit risk model prediction apparatus 300 shown in fig. 3, the apparatus 300 includes an obtaining module 301, a screening module 302 and a prediction module 303.

An obtaining module 301, configured to obtain to-be-processed data, where the to-be-processed data includes financial data of a customer;

a screening module 302, configured to input the data to be processed into a screening model to obtain screening data; the screening model comprises a penalty function, and the penalty function is used for removing part of the data to be processed;

a prediction module 303 for performing credit risk model prediction using the screening data.

As a possible implementation, the apparatus 300 further includes:

As a possible implementation, the screening model includes: a LASSO regression model that includes the penalty function.

As a possible implementation, the prediction module 303 is configured to perform credit risk model prediction on the screening data by using a logistic regression algorithm.

As a possible implementation, the apparatus 300 further includes:

a result module to obtain a result of the credit risk model prediction;

The embodiment of the application also provides corresponding equipment and a computer storage medium, which are used for realizing the scheme provided by the embodiment of the application.

Wherein the apparatus comprises a memory for storing instructions or code and a processor for executing the instructions or code to cause the apparatus to perform a credit risk model prediction method as described in any embodiment of the present application.

The computer storage medium has code stored therein that when executed, an apparatus executing the code implements a credit risk model prediction method as described in any embodiment of the present application.

In practice, the computer readable storage medium may take any combination of one or more computer readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present embodiment, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the apparatus embodiments are substantially similar to the method embodiments and therefore are described in a relatively simple manner, and reference may be made to some of the description of the method embodiments for relevant points. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement without inventive effort.

The above description is only an exemplary embodiment of the present application, and is not intended to limit the scope of the present application.

Claims

1. A credit risk model prediction method, the method comprising:

credit risk model prediction is performed using the screening data.

2. The method of claim 1, wherein before inputting the data to be processed into a screening model to obtain screened data, the method further comprises:

3. The method of claim 1, wherein the screening the model comprises: a LASSO regression model including the penalty function.

4. The method according to claim 1, wherein the using the screening data for credit risk model prediction comprises:

5. The method of claim 1, further comprising:

obtaining results of the credit risk model prediction;

converting the results to a score that reflects a condition predicted by the credit risk model;

6. A credit risk model prediction apparatus, the apparatus comprising:

7. The apparatus of claim 6, further comprising:

8. The apparatus of claim 6, further comprising:

a result module to obtain a result of the credit risk model prediction;

9. An apparatus, characterized in that the apparatus comprises a memory for storing instructions or code and a processor for executing the instructions or code to cause the apparatus to perform the credit risk model prediction method of any one of claims 1 to 5.

10. A computer storage medium having code stored therein, the computer storage device executing the code implementing the credit risk model prediction method of any one of claims 1 to 5 when the code is executed.