Credit risk prediction method and system
Through grouping and multivariate linear regression models, the proportion of mob12 60+ overdue amounts of credit customers is predicted, combined with the vintage curve and annualized loss rate prediction, the problem of lack of objectivity and consistency in the existing credit risk prediction methods is solved, and more efficient and accurate risk management.
Patent Information
- Application Number
- CN202411527273.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-30
- Publication Date
- 2025-02-21
AI Technical Summary
The existing credit risk prediction methods rely on the experience and subjective judgment of senior experts, lack objectivity and consistency, which affects the accuracy of prediction and the efficiency of risk management.
By obtaining the identity information and loan information of customers, the customers are divided into multiple customer groups, and the pre-trained multivariate linear regression model is used to predict the proportion of mob12 60+ overdue amounts of each customer group, combining the vintage curve estimate and year To provide a more comprehensive risk assessment.
It significantly improves the accuracy and objectivity of credit risk prediction, reduces the impact of personal experience and subjective judgment, and achieves more refined risk management and more efficient risk assessment.
Smart Images
Figure CN119494720A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of risk control decision prediction and analysis, and in particular to a credit risk prediction method and system. Background Art
[0002] In the financial sector, risk prediction and management are crucial links. Credit risk not only affects the asset security of financial institutions, but also directly affects their profitability and market competitiveness. In the consumer credit business, accurately estimating the annualized loss rate (annual loss) of loans is crucial for financial institutions' risk management, pricing strategies, and profit forecasts. Annual loss estimation usually involves analyzing the performance of loan customer groups based on historical data, and then predicting possible future losses.
[0003] At present, the existing credit risk prediction methods mainly rely on the experience and judgment of senior credit risk experts, combined with the market environment and historical data to estimate loan losses. Although this method can take into account a variety of factors, it is greatly influenced by personal experience and subjective judgment, and lacks objectivity and consistency. This not only affects the accuracy of the prediction, but also reduces the efficiency of risk management. Summary of the invention
[0004] In view of this, it is necessary to provide a credit risk prediction method and system to solve the above-mentioned defects of the prior art.
[0005] In order to solve the above problems, in a first aspect, an embodiment of the present invention provides a credit risk prediction method, comprising:
[0006] Obtain the customer's identity information and loan information; the loan information includes the number of terms, pricing, and overdue information of the loan;
[0007] Divide customers into multiple customer groups based on their basic information and overdue information on IOUs;
[0008] For each customer group, based on the loan information within three periods, the pre-trained machine learning model is used to predict the proportion of overdue amounts of mob12 60+ in the customer group;
[0009] Among them, mob represents the age of the account; the overdue amount ratio of mob12 60+ represents the ratio of the overdue amount with more than 60 days overdue to the total loan amount after the end of the 12th period of the loan note.
[0010] Preferably, after predicting the mob12 60+ overdue amount ratio of each customer group by a pre-trained machine learning model based on the loan information within three periods, the method further comprises:
[0011] The predicted results of the overdue amount ratio of mob12 60+ for each customer group are weighted and summed to obtain the predicted value of the overdue amount ratio of mob12 60+ for all customers.
[0012] Preferably, the machine learning model is a multivariate linear regression model.
[0013] Preferably, before predicting the proportion of overdue amount of mob12 60+ of the customer group by a pre-trained machine learning model based on the loan information within three periods, the method further includes:
[0014] Based on the customer's identity information and loan information, a data set for a multivariate linear regression model is constructed, and the data set is divided into a training set and a test set according to a preset ratio;
[0015] Using the training set data, the model parameters are fitted by the multivariate linear regression algorithm;
[0016] Evaluate the prediction performance of the model on the test set and calculate the error between the predicted value and the actual value;
[0017] The model parameters are adjusted according to the error results, and finally a trained multivariate linear regression model is obtained.
[0018] Preferably, after dividing the customers into multiple customer groups according to their basic information and overdue information of the loan notes, the method further comprises:
[0019] For each customer group, estimate the vintage curve of the customer group based on the loan information within three periods;
[0020] Summarize and integrate the vintage curves of each customer group and calculate the estimated annualized loss rate.
[0021] In a second aspect, an embodiment of the present invention provides a credit risk prediction system, comprising:
[0022] The data acquisition module is used to obtain the customer's identity information and loan information; wherein the loan information includes the number of terms, pricing and overdue information of the loan;
[0023] The customer group classification module is used to classify customers into multiple customer groups based on their basic information and overdue information of IOUs;
[0024] The risk prediction module is used to predict the percentage of overdue amounts of mob12 60+ for each customer group based on the loan information within three periods through a pre-trained machine learning model;
[0025] Among them, mob represents the age of the account; the overdue amount ratio of mob12 60+ represents the ratio of the overdue amount with more than 60 days overdue to the total loan amount after the end of the 12th period of the loan note.
[0026] Preferably, the system further comprises:
[0027] The annual loss estimation module is used to estimate the vintage curve of each customer group based on the loan information within three periods; the vintage curves of each customer group are summarized and integrated to calculate the estimated value of the annualized loss rate.
[0028] In a third aspect, the present invention further provides an electronic device, including a memory and a processor, wherein:
[0029] The memory is used to store programs;
[0030] The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the credit risk prediction method as described in the embodiment of the first aspect of the present invention.
[0031] In a fourth aspect, the present invention further provides a computer-readable storage medium for storing computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the credit risk prediction method as described in the embodiment of the first aspect of the present invention.
[0032] Compared with the prior art, the credit risk prediction method and system provided by the present invention have the following beneficial effects:
[0033] 1) The present invention predicts the proportion of overdue amount of mob12 60+ of customers through machine learning model, which can predict credit risk more objectively and accurately, significantly improve the accuracy of prediction, and reduce the influence of personal experience and subjective judgment on the prediction results.
[0034] 2) The present invention achieves more refined risk management by segmenting customer groups and making predictions for each customer group, which helps credit institutions better understand and control the risks of different customer groups. In addition, the present invention also combines vintage curve estimation and annualized loss rate prediction to provide a more comprehensive risk assessment perspective, further improving the efficiency and effectiveness of risk management. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 Schematic diagram of risk control decision log analysis scenario provided by existing technology;
[0036] Figure 2 A schematic diagram of the prediction results of the overdue amount ratio of mob12 60+ by the multivariate linear regression model provided by the present invention;
[0037] Figure 3 This is a structural block diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0038] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.
[0039] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0040] At present, the existing credit risk prediction methods mainly rely on the experience and judgment of senior credit risk experts, combined with the market environment and historical data to estimate loan losses. Although this method can take into account a variety of factors, it is greatly influenced by personal experience and subjective judgment, and lacks objectivity and consistency. This not only affects the accuracy of the prediction, but also reduces the efficiency of risk management.
[0041] In view of this, the present invention provides a credit risk prediction method, which predicts the proportion of overdue amount of mob12 60+ of customers through machine learning model, and can predict credit risk more objectively and accurately, significantly improve the accuracy of prediction, and reduce the influence of personal experience and subjective judgment on the prediction results. The following will be explained and introduced through multiple embodiments.
[0042] Figure 1 The credit risk prediction method provided by the present invention is shown in the flowchart. Figure 1 The credit risk prediction method provided by the present invention at least comprises the following steps:
[0043] Step S1, obtaining the customer's identity information and loan note information; wherein the loan note information includes the number of terms, pricing and overdue information of the loan note.
[0044] Specifically, obtain the IOU from the system and ensure that each IOU has completed 12 periods (mob12). Extract the IOU information, including but not limited to the number of periods, pricing, and overdue information. Extract the basic information of the corresponding customer, such as age, education, etc.
[0045] In this embodiment, the overdue information of the loan note includes fpd1-fpd30 (overdue situation 1-30 days after the first overdue), mob115+ (the proportion of overdue amount more than 15 days after mob1), mob2 30+ to mob12 30+ (the proportion of overdue amount more than 30 days each month from mob2 to mob12), mob1260+ (the proportion of overdue amount more than 60 days after mob12), etc.
[0046] Among them, mob (month of book) represents the age of the account; mob1 15+ represents the proportion of loans that are overdue for more than 15 days after the end of the first full repayment month (mob1) of the loan note (loan).
[0047] "mob12 60+" refers to loans or credit products that are overdue for more than 60 days when the loan note has been aged for 12 periods. It is used to measure the overdue risk and quality of credit products at the end of their life cycle. The prediction of mob12 60+ has important risk control significance for financial institutions.
[0048] Step S2, dividing the customers into multiple customer groups according to their basic information and overdue information of the loan notes.
[0049] Specifically, customers are divided into different groups based on their basic information (such as age, education) and overdue information of the loan notes (such as mob1 15+, mob2 30+, etc.). These groups should show certain differences in overdue behavior and risk characteristics to facilitate subsequent accurate analysis and prediction. In addition, the repayment performance of the divided customer groups is stable, and the performance of the same customer group should not be significantly different at different loan periods. The purpose of grouping is to improve the accuracy and pertinence of predictions.
[0050] In actual operation, the number of each customer group shall not be less than 2% of the total number of customers. Customers can be divided into the following 10 customer groups according to age and IOU pricing:
[0051] Age>26and<=30, price>15 to <=19
[0052] Age>30, price>15 to <=19
[0053] Age>26and<=30, price>10 to <=15
[0054] Age <= 26, price > 10 to <= 15
[0055] Age <= 26, price > 15 to <= 19
[0056] Age>30, Price>19 and above
[0057] Age <= 26, price > 19 and above
[0058] Age>30, price>10 to <=15
[0059] Age>26and<=30, price>19 and above
[0060] Here, pricing refers to the annualized interest rate of the loan note, and a pricing of 15 means an annualized 15%.
[0061] In this embodiment, by segmenting customer groups and making predictions for each customer group, more refined risk management is achieved, which helps credit institutions better understand and control the risks of different customer groups.
[0062] Step S3, for each customer group, based on the loan information within three periods, predict the proportion of overdue amount of mob12 60+ of the customer group through the pre-trained machine learning model;
[0063] In this embodiment, a single customer group is analyzed, and the analysis method can use curve fitting, polynomial regression and multivariate regression. Finally, after comparative experiments of various analysis and prediction methods, the embodiment of the present invention uses a multivariate linear regression model to predict the proportion of overdue amount of mob12 60+ of the customer group.
[0064] In a preferred embodiment of the present invention, it is understandable that before using the multiple linear regression model to predict the proportion of overdue amount of mob12 60+ of the customer group, the multiple linear regression model needs to be trained. In this embodiment, the training process of the multiple linear regression model includes:
[0065] S31, based on the customer's identity information and loan information, construct a data set for a multivariate linear regression model, and divide the data set into a training set and a test set according to a preset ratio (eg, 7:3).
[0066] S32, using the training set data, fit the model parameters through the multivariate linear regression algorithm.
[0067] Construct a multiple linear regression model, the model form is:
[0068] Y=β0+β1X1+β2X2+…+β n X n +ε
[0069] Where Y is the percentage of overdue amount of mob12 60+, X1, X2, ..., X n For the selected features, β0, β1, β2, ..., β n is the regression coefficient and ε is the error term.
[0070] Define a loss function. In this embodiment, mean square error (MSE) can be used as the loss function.
[0071] Use an optimization algorithm such as least squares or gradient descent to minimize the loss function to solve the regression coefficients β0, β1, β2, ..., β n .
[0072] S33, evaluate the prediction performance of the model on the test set and calculate the error between the predicted value and the actual value, that is, the mean square error MSE.
[0073] S34, adjusting the model parameters according to the error results to improve the prediction performance of the model, and finally obtaining a trained multivariate linear regression model.
[0074] In step S3, for each customer group, the loan information within the three periods of the customer group is input into the trained multiple linear regression model, and the multiple linear regression model outputs the predicted value of the overdue amount ratio of mob12 60+ of the customer group. Among them, the loan information within the three periods includes (the overdue amount ratio of 1-30 days after the first overdue), mob1 15+ (the overdue amount ratio of more than 15 days after the first period of the loan), mob215+ (the overdue amount ratio of more than 15 days after the end of the second period of the loan) and mob230+ (the overdue amount ratio of more than 30 days after the end of the second period of the loan).
[0075] In this example, the data within three periods (including fpd1-fpd30, mob1 15+, mob215+, mob2 30+) are used as input features of the multiple linear regression model to predict the value of mob12 60+. The prediction results are as follows: Figure 2 As shown, the horizontal axis in the figure is the number of periods of the loan note, and the vertical axis is the proportion of overdue amounts of mob1260+. The multivariate linear regression model provided by the present invention can predict the risk status of the loan note in a longer period when there is less information about the loan note (no more than three periods). Finally, the model outputs the predicted values of the proportion of overdue amounts of mob12 60+ for ten customer groups, which are 0.01540654, 0.0208096, 0.00864152, 0.0056405, 0.00903529, 0.03201662, 0.01463384, 0.01793732, 0.01162087, and 0.02494019 respectively.
[0076] Furthermore, the predicted results of the mob12 60+ overdue amount ratio of each customer group are weighted and summed to obtain the predicted value of the mob12 60+ overdue amount ratio of all customers. In this embodiment, the predicted results of the mob12 60+ overdue amount ratio of the above ten customer groups are weighted and summed to obtain the predicted value of the mob12 60+ overdue amount ratio of all customers, which is 0.017333.
[0077] The present invention predicts the proportion of overdue amounts of mob12 60+ of customers through a machine learning model, which can more objectively and accurately predict credit risks, significantly improve the accuracy of prediction, and reduce the influence of personal experience and subjective judgment on the prediction results.
[0078] In a preferred embodiment of the present invention, the embodiment of the present invention predicts the proportion of overdue amount of mob12 60+ of the customer group through a pre-trained machine learning model based on the mob1 loan information.
[0079] It is understandable that if mob2 and previous data are used to predict credit risk (the proportion of overdue amounts of mob12 60+), it will take three natural months after the loan is issued to obtain the prediction result, resulting in a lag in the prediction result. Therefore, this embodiment proposes a mob1 prediction scheme to improve the lag problem that exists when using mob2 data (i.e., aging data of the second month after the loan is issued) for prediction in existing credit risk prediction technologies. Specifically, the core of this optimization scheme is to reduce the time range of data use and only use mob1 (aging data of the first month after the loan is issued) and previous data for prediction, thereby significantly improving the timeliness of the prediction.
[0080] In a preferred embodiment of the present invention, after dividing the customers into multiple customer groups according to the basic information of the customers and the overdue information of the loan notes in step S2, the method further comprises:
[0081] For each customer group, estimate the vintage curve of the customer group based on the loan information within three periods; summarize and integrate the vintage curves of each customer group to calculate the estimated annualized loss rate.
[0082] Specifically, the vintage overdue rate is an important indicator in credit risk management, used to assess the quality of loan assets in a specific period. Vintage overdue rate = overdue amount of a specific age / total loan amount in the base month.
[0083] The vintage curve is a chart showing the change in the overdue rate of loans issued in different time periods over time, which can help understand the risk performance of loans in different time periods. In this embodiment, the pre-trained multivariate regression prediction model can capture the trend of customer overdue behavior based on the loan information within 3 periods, so as to more accurately estimate the vintage curve of each customer group.
[0084] Furthermore, after obtaining the vintage curve forecast for each customer group, these results are summarized and integrated. This usually involves weighted averaging or other forms of comprehensive processing of the forecast results of different customer groups to obtain an overall vintage curve forecast. This overall vintage curve will reflect the overdue behavior trend and risk performance of the entire customer group.
[0085] Finally, based on the aggregated vintage curve estimates, the estimated annualized loss rate is calculated. The annualized loss rate is a measure of the average loss rate that a loan or portfolio may suffer within a year.
[0086] The present invention combines the mob12 60+ prediction and the annualized loss rate prediction scheme, provides a more comprehensive risk assessment perspective, and further improves the efficiency and effectiveness of risk management. In actual operation, the mob12 60+ prediction results and the annualized loss rate prediction results are shown in Table 1.
[0087] Table 1
[0088]
[0089] In Table 1, the numerical mob1 forecast represents the value obtained by using the data set from the first overdue to mob1 through the multivariate linear regression model. This reflects the impact of shorter-term data on the forecast results.
[0090] The numerical mob2 forecast represents the value predicted by the multivariate linear regression model based on the longer time period data set from the first overdue to mob2. The risk forecast represents the prediction results of a specific risk model for these two indicators.
[0091] In a preferred embodiment of the present invention, the present invention provides a credit risk prediction system, the system comprising:
[0092] The data acquisition module is used to obtain the customer's identity information and loan information; wherein the loan information includes the number of terms, pricing and overdue information of the loan;
[0093] The customer group classification module is used to classify customers into multiple customer groups based on their basic information and overdue information of IOUs;
[0094] The risk prediction module is used to predict the percentage of overdue amounts of mob12 60+ for each customer group based on the loan information within three periods through a pre-trained machine learning model;
[0095] Among them, mob represents the age of the account; the overdue amount ratio of mob12 60+ represents the ratio of the overdue amount with more than 60 days overdue to the total loan amount after the end of the 12th period of the loan note.
[0096] The annual loss estimation module is used to estimate the vintage curve of each customer group based on the loan information within three periods; the vintage curves of each customer group are summarized and integrated to calculate the estimated value of the annualized loss rate.
[0097] The credit risk prediction system provided by the present invention executes the credit risk prediction methods provided by the aforementioned embodiments through the aforementioned modules. The credit risk prediction methods have been described in detail in the aforementioned embodiments, and will not be repeated in this embodiment.
[0098] Figure 3 The structural block diagram of the electronic device provided by the present invention is as follows: Figure 3 As shown, the present invention also provides an electronic device, the electronic device 300 can be a computing device such as a mobile terminal, a desktop computer, a notebook, a palm computer, a server, etc. The electronic device 300 includes a processor 301 and a memory 302, wherein the memory 302 stores a credit risk prediction program 303.
[0099] In some embodiments, the memory 302 may be an internal storage unit of a computer device, such as a hard disk or memory of the computer device. In other embodiments, the memory 302 may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device. Furthermore, the memory 302 may also include both an internal storage unit of the computer device and an external storage device. The memory 302 is used to store application software and various types of data installed on the computer device, such as program codes installed on the computer device. The memory 302 may also be used to temporarily store data that has been output or is to be output. In one embodiment, when the credit risk prediction program 303 is executed by the processor 301, the following steps are implemented:
[0100] Obtain the customer's identity information and loan information; the loan information includes the number of terms, pricing, and overdue information of the loan;
[0101] Divide customers into multiple customer groups based on their basic information and overdue information on IOUs;
[0102] For each customer group, based on the loan information within three periods, the pre-trained machine learning model is used to predict the proportion of overdue amounts of mob12 60+ of the customer group.
[0103] In some embodiments, the processor 301 may be a central processing unit (CPU), a microprocessor or other data processing chip, which is used to run the program code or process data stored in the memory 302, such as executing a credit risk prediction program.
[0104] This embodiment also provides a computer-readable storage medium on which a credit risk prediction program is stored. When the credit risk prediction program is executed by a processor, the following steps are implemented:
[0105] Obtain the customer's identity information and loan information; the loan information includes the number of terms, pricing, and overdue information of the loan;
[0106] Divide customers into multiple customer groups based on their basic information and overdue information on IOUs;
[0107] For each customer group, based on the loan information within three periods, the pre-trained machine learning model is used to predict the proportion of overdue amounts of mob12 60+ of the customer group.
[0108] The above-described embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.
[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A credit risk prediction method, characterized in that: include: Obtain the customer's identity information and loan information; the loan information includes the number of terms, pricing, and overdue information of the loan; Divide customers into multiple customer groups based on their basic information and overdue information on IOUs; For each customer group, based on the loan information within three periods, the pre-trained machine learning model is used to predict the proportion of overdue amounts of mob12 60+ in the customer group; Among them, mob represents the age of the account; the overdue amount ratio of mob12 60+ represents the ratio of the overdue amount with more than 60 days overdue to the total loan amount after the end of the 12th period of the loan note.
2. The credit risk prediction method according to claim 1, characterized in that: After predicting the mob12 60+ overdue amount ratio of each customer group based on the loan information in three periods by using a pre-trained machine learning model, the method further includes: The predicted results of the overdue amount ratio of mob12 60+ for each customer group are weighted and summed to obtain the predicted value of the overdue amount ratio of mob12 60+ for all customers.
3. The credit risk prediction method according to claim 1, characterized in that: The machine learning model is a multiple linear regression model.
4. The credit risk prediction method according to claim 3, characterized in that: Before predicting the mob12 60+ overdue amount ratio of the customer group through a pre-trained machine learning model based on the loan information within three periods, the method further includes: Based on the customer's identity information and loan information, a data set for a multivariate linear regression model is constructed, and the data set is divided into a training set and a test set according to a preset ratio; Using the training set data, the model parameters are fitted by the multivariate linear regression algorithm; Evaluate the prediction performance of the model on the test set and calculate the error between the predicted value and the actual value; The model parameters are adjusted according to the error results, and finally a trained multivariate linear regression model is obtained.
5. The credit risk prediction method according to claim 1, characterized in that: After dividing the customers into multiple customer groups according to their basic information and overdue information of the loan notes, the method further includes: For each customer group, estimate the vintage curve of the customer group based on the loan information within three periods; Summarize and integrate the vintage curves of each customer group and calculate the estimated annualized loss rate.
6. A credit risk prediction system applied to the credit risk prediction method according to any one of claims 1 to 5, characterized in that: include: The data acquisition module is used to obtain the customer's identity information and loan information; wherein the loan information includes the number of terms, pricing and overdue information of the loan; The customer group classification module is used to classify customers into multiple customer groups based on their basic information and overdue information of IOUs; The risk prediction module is used to predict the percentage of overdue amounts of mob12 60+ for each customer group based on the loan information within three periods through a pre-trained machine learning model; Among them, mob represents the age of the account; the overdue amount ratio of mob12 60+ represents the ratio of the overdue amount with more than 60 days overdue to the total loan amount after the end of the 12th period of the loan note.
7. The credit risk prediction system according to claim 6, characterized in that: Also includes: The annual loss estimation module is used to estimate the vintage curve of each customer group based on the loan information within three periods; Summarize and integrate the vintage curves of each customer group and calculate the estimated annualized loss rate.
8. An electronic device, It is characterized in that comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the credit risk prediction method described in any one of claims 1 to 5 above.
9. A computer-readable storage medium, characterized in that: Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps of the credit risk prediction method described in any one of claims 1 to 5 above.