Risk prediction method and system based on customer structure change

Abstract

The invention relates to the technical field of consumer credit risk management, in particular to a customer structure change-based risk prediction method and system, and the method comprises the steps: constructing analysis sample data; performing binning processing on the analysis sample data according to the risk data by selecting two credit risk models; constructing an original matrix; constructing a client hierarchical division analysis result; customer structure analysis is carried out on customers without risk performance; and carrying out customer structure analysis on the built-in customer. According to the method, the two credit risk models with the KS values larger than or equal to the preset threshold value are selected to serve as the basis of customer hierarchical division analysis processing, high-precision grade division is conducted on customer structures through the risk data of historical customers by means of the binning technology, the matrix analysis technology and the customer hierarchical division analysis processing technology, and the customer hierarchical division analysis processing efficiency is improved. And the customer hierarchical division analysis result is obtained and is applied to risk prediction of customers without risk performance and imbedded customers, so that the accuracy of risk prediction is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of consumer credit risk management, and in particular to a risk prediction method and system based on changes in customer structure. Background Art

[0002] In recent years, with the continuous development of Internet technology, my country's consumer credit business has developed rapidly. With the growth and fluctuation of the number of customers and loan amount, the default of loan users is crucial to business development, and it is necessary to timely and effectively predict the risks of users in order to maximize benefits. Existing risk prediction strategies usually use historical samples and credit risk models for risk prediction. However, there may be large differences between historical samples and current samples, which will affect the accuracy of risk prediction. In addition, risk performance has a certain lag. If the actual risk occurs and it is found that there is a large difference with the risk prediction results, it may lead to greater risk losses.

[0003] To this end, the present invention provides a risk prediction method and system based on changes in customer structure. Summary of the invention

[0004] Based on this, it is necessary to provide a risk prediction method and system based on changes in customer structure to address the above technical issues. This method aims to solve the problems of poor accuracy and large lag in existing risk prediction methods.

[0005] According to one aspect of the present invention, a risk prediction method based on changes in customer structure is provided, including: obtaining historical customers and their risk data, and constructing analysis sample data; selecting two credit risk models with KS values ​​greater than or equal to a preset threshold, and binning the analysis sample data according to the risk data using the two selected credit risk models to obtain a first binning result and a second binning result; using the first binning result and the second binning result as rows and columns, respectively, to construct an original matrix, and calculate the customer quantity ratio and risk data of each grid in the original matrix; merging the grids in the original matrix according to a preset risk data classification level rule to obtain a merged result based on a group dimension, and constructing a customer stratification analysis result, as well as obtaining a mapping relationship between the customer stratification level and the model score of the two credit risk models; based on the customer stratification analysis result, performing a customer structure analysis on customers with no risk performance to obtain a risk prediction result for customers with no risk performance; in the iterative impact assessment process of the risk strategy, based on the customer stratification analysis result, performing a customer structure analysis on the placed customers to obtain a risk prediction result for the placed customers.

[0006] In some optional implementations of some embodiments, the analysis sample set is binned according to the risk data using the two selected credit risk models to obtain a first binning result and a second binning result, including: the analysis sample set is binned according to the risk data using the two selected credit risk models, wherein the binning adopts an equal frequency binning method or a chi-square binning method to obtain a first binning result and a second binning result including a model score, number of customers, quantity proportion, number of overdue payments, and risk value.

[0007] In some optional implementations of some embodiments, the first binning results and the second binning results are used as rows and columns respectively to construct the original matrix, and the customer quantity ratio and risk data of each grid in the original matrix are calculated, including: according to the four group dimensions of customer number, quantity ratio, overdue number and risk value, the model scores of the first binning results and the model scores of the second binning results are used as rows and columns respectively to construct a customer number matrix, a quantity ratio matrix, an overdue number matrix and a risk value matrix respectively, and calculate the customer quantity ratio of each grid in the quantity ratio matrix, and calculate the risk data of each grid in the risk value matrix.

[0008] In some optional implementations of some embodiments, the grids in the original matrix are merged according to the preset risk data classification level rules to obtain a merged result based on the group dimension, and a customer stratification analysis result is constructed, as well as a mapping relationship between the customer stratification level and the model scores of the two credit risk models, including: based on the preset risk data classification level rules, the grids with similar risk values ​​in the risk value matrix are merged to obtain a merged result of the corresponding risk data; the grids in the customer number matrix, the quantity proportion matrix, the overdue number matrix and the risk value matrix are merged according to the grids in the customer number matrix, the quantity proportion matrix, the overdue number matrix and the risk value matrix. According to the mapping relationship of the grids, the grids within the customer number matrix, the quantity proportion matrix, and the overdue number matrix are merged respectively to obtain the merged results of the corresponding customer number, the merged results of the corresponding quantity proportion, and the merged results of the corresponding overdue number; based on the mapping relationship between the risk data classification level and the customer stratification classification level, according to the customer stratification classification level, the merged results of the corresponding risk data, the merged results of the corresponding customer number, the merged results of the corresponding quantity proportion, and the merged results of the corresponding overdue number, the customer stratification classification analysis results are constructed, and the mapping relationship between the customer stratification classification level and the model scores of the two credit risk models is obtained.

[0009] In some optional implementations of some embodiments, the method of performing a customer structure analysis on customers with no risk performance based on the customer stratification analysis results to obtain risk prediction results for customers with no risk performance includes: scoring customers with no risk performance using two selected credit risk models to obtain two model scores for customers with no risk performance, and grading customers with no risk performance according to a mapping relationship between customer stratification grades and model scores of the two credit risk models to obtain a first customer stratification grade result; comparing the first customer stratification grade result with the customer stratification analysis result to obtain a first customer structure change identification result; performing a risk estimation analysis on new loans to customers with no risk performance based on the customer stratification analysis result, calculating the product of the proportion of the number of customers at each customer stratification grade in the first customer stratification grade result and the risk value of the analysis sample data at the corresponding customer stratification grade, and summing the products obtained at all customer stratification grades to obtain a risk prediction result for customers with no risk performance.

[0010] In some optional implementations of some embodiments, during the iterative impact assessment of the risk strategy, based on the customer stratification results, a customer structure analysis is performed on the placed customers to obtain risk prediction results corresponding to the placed customers, including: during the iterative impact assessment of the risk strategy, the placed customers are scored by the two selected credit risk models to obtain two model scores corresponding to the placed customers, and the placed customers are graded according to the mapping relationship between the customer stratification levels and the model scores of the two credit risk models to obtain a second customer stratification level result; based on the customer stratification analysis results, the product of the proportion of the number of customers at each customer stratification level in the second customer stratification level result and the risk value of the analysis sample data at the corresponding customer stratification level is calculated, and the products obtained at all customer stratification levels are summed to obtain the risk prediction results corresponding to the placed customers.

[0011] In some optional implementation methods of some embodiments, it also includes iterative updating of customer stratification analysis results, including: determining whether the credit risk model currently in use has risk identification; if the credit risk model currently in use does not have risk identification, then reselecting a credit risk model with a KS value greater than or equal to a preset threshold to replace the credit risk model without risk identification, and re-processing the customer stratification level of the analysis sample data to obtain updated customer stratification analysis results; if the credit risk model currently in use has risk identification, but the risk identification effect is average, then adding a credit risk model with a KS value greater than or equal to the preset threshold, adjusting the preset risk data classification level rules, and re-merging the grids inside the original matrix to obtain updated customer stratification analysis results.

[0012] According to a second aspect of the present invention, a risk prediction system based on customer structure changes is provided, including: a data acquisition module, used to acquire historical customers and their risk data, and construct analysis sample data; a binning processing module, used to select two credit risk models with KS values ​​greater than or equal to a preset threshold, and bin the analysis sample data according to the risk data using the selected two credit risk models to obtain a first binning result and a second binning result; a matrix construction module, used to use the first binning result and the second binning result as rows and columns, respectively, to construct an original matrix, and calculate the customer quantity ratio and risk data of each grid in the original matrix; a matrix processing module, used to divide the risk data according to the preset risk data, etc. The grids inside the original matrix are merged according to the level rule to obtain the merged result based on the group dimension, and the customer stratification analysis result is constructed, as well as the mapping relationship between the customer stratification level and the model score of the two credit risk models; the risk prediction module is used to perform customer structure analysis on the placed customers based on the customer stratification analysis result in the iterative impact assessment process of the risk strategy to obtain the risk prediction result of the corresponding placed customers, and to perform customer structure analysis on the placed customers based on the customer stratification analysis result in the iterative impact assessment process of the risk strategy to obtain the risk prediction result of the corresponding placed customers; the update module is used to iteratively update the customer stratification analysis result.

[0013] According to a third aspect of the present invention, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0014] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium storing a computer program, wherein the computer program implements the steps of the above method when executed by a processor.

[0015] The advantages and beneficial effects of the present invention are as follows: the risk prediction method and system based on customer structure changes provided by the present invention select two credit risk models with KS values ​​greater than or equal to a preset threshold as the basis for customer stratification analysis and processing, and use binning technology, matrix analysis technology and customer stratification analysis and processing technology to achieve high-precision grading of customer structure using historical customer risk data, obtain customer stratification analysis results, and based on the customer stratification analysis results, perform risk prediction on customers with no risk performance and placed customers, so as to improve the accuracy of risk prediction and avoid the problem of large lag. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Flow chart of a risk prediction method based on changes in customer structure in an embodiment of the present invention.

[0017] Figure 2 Schematic diagram of a risk prediction system based on changes in customer structure in an embodiment of the present invention.

[0018] Figure 3 Schematic diagram of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below through specific implementation methods in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] Embodiment 1

[0021] Reference Figure 1 , a risk prediction method based on changes in customer structure, the method specifically includes the following steps.

[0022] Step S1: Obtain historical customer and risk data and construct analysis sample data.

[0023] Step S2: Select two credit risk models whose KS values ​​are greater than or equal to a preset threshold, and use the two selected credit risk models to bin the analysis sample data according to the risk data to obtain a first binning result and a second binning result.

[0024] In some embodiments, various types of credit risk models are used in daily risk control strategies. Credit risk models include internal company models or external tripartite models. Two credit risk models with better monotonically sorted risk data are selected as the basis for customer stratification analysis and processing. It should be noted that the KS value (Kolmogorov-Smirnov value) of the credit risk model is used to measure the ability of the credit risk model to distinguish between good and bad customers. The larger the KS value, the better the ranking of the credit risk model. Among them, the KS value of a credit risk model with better ranking is usually between 0.2-0.7. In addition, the preset threshold can be set according to actual conditions. For example, the preset threshold is 0.3.

[0025] In some embodiments, the analysis sample set is binned according to the risk data using two selected credit risk models to obtain a first binning result and a second binning result, including: the analysis sample set is binned according to the risk data using two selected credit risk models, wherein the binning adopts an equal frequency binning method or a chi-square binning method to obtain a first binning result and a second binning result including model score, number of customers, number percentage, number of overdue payments, and risk value, respectively; wherein the customer number percentage is the ratio of the number of customers to the total number, the risk data is the number of overdue payments / total number, and the total part in the risk data refers to the average risk of this group of people.

[0026] In some embodiments, the analysis sample sets are binned according to the risk data of the two selected credit risk models. The equal frequency binning method or the chi-square binning method can be used. The binning results using the equal frequency binning method are shown in Table 1, and the binning results using the chi-square binning method are shown in Table 2.

[0027] Table 1 Results of the first binning

[0028] Model points Number of customers Quantity ratio Overdue Risk Value 0-20 5000 20％ 195 3.90％ 21-45 5000 20％ 180 3.60％ 46-60 5000 20％ 125 2.50％ 61-85 5000 20％ 100 2.00％ 86-100 5000 20％ 70 1.40％ total 25000 100 670 2.68％

[0029] Table 2 Second binning results

[0030] Model points Number of customers Quantity ratio Overdue Risk Value 0-15 3000 12％ 119 4.00％ 16-30 4500 18％ 160 3.60％ 31-45 3750 15％ 119 3.20％ 46-65 5000 20％ 125 2.50％ 66-85 5000 20％ 103 2.10％ 86-100 3750 15％ 44 1.20％ total 25000 100 670 2.68％

[0031] Step S3: Use the first binning results and the second binning results as rows and columns respectively, construct the original matrix, and calculate the customer quantity ratio and risk data of each grid in the original matrix.

[0032] In some embodiments, the first binning results and the second binning results are used as rows and columns respectively to construct the original matrix, and the customer quantity ratio and risk data of each grid in the original matrix are calculated, including: according to the four group dimensions of customer number, quantity ratio, overdue number and risk value, the model scores of the first binning results and the model scores of the second binning results are used as rows and columns respectively to construct a customer number matrix, a quantity ratio matrix, an overdue number matrix and a risk value matrix respectively, and the customer quantity ratio of each grid in the quantity ratio matrix is ​​calculated, and the risk data of each grid in the risk value matrix is ​​calculated.

[0033] In some embodiments, according to the four grouping dimensions of customer number, quantity proportion, overdue number and risk value, the first binning results and the second binning results are used as rows and columns to form an N*M matrix respectively, and the customer quantity proportion of each grid in the quantity proportion matrix is ​​calculated, and the risk data of each grid in the risk value matrix is ​​calculated. The constructed customer number matrix, quantity proportion matrix, overdue number matrix and risk value matrix are shown in Table 3-6.

[0034] Table 3 Customer number matrix

[0035]

[0036] Table 4 Quantity ratio matrix

[0037]

[0038] Note: Examples of data percentage calculations are: 500 / 25000*100%=2%, 3000 / 25000*100%=12%.

[0039] Table 5 Overdue number matrix

[0040]

[0041]

[0042] Table 6 Risk value matrix

[0043]

[0044] Note: Examples of risk value calculations are: 29 / 500*100%=5.8%, 119 / 3000*100%=4%, 670 / 25000*100%=2.7%.

[0045] Step S4: According to the preset risk data classification rules, the grids in the original matrix are merged to obtain the merging results based on the group dimension, and the customer stratification analysis results are constructed, as well as the mapping relationship between the customer stratification levels and the model scores of the two credit risk models.

[0046] In some embodiments, according to the preset risk data classification level rules, the grids in the original matrix are merged to obtain the merged result based on the group dimension, and the customer stratification analysis results are constructed, and the mapping relationship between the customer stratification level and the model scores of the two credit risk models is obtained, including: based on the preset risk data classification level rules, the grids with similar risk values ​​in the risk value matrix are merged to obtain the merged result of the corresponding risk data; according to the mapping relationship between the grids in the customer number matrix, the quantity proportion matrix, the overdue number matrix and the grids in the risk value matrix, the customer number matrix, the quantity proportion matrix and the grids in the risk value matrix are merged. , the grids inside the overdue number matrix are merged respectively to obtain the merged results of the corresponding customer numbers, the merged results of the corresponding number percentages and the merged results of the corresponding overdue numbers; based on the mapping relationship between risk data classification levels and customer stratification classification levels, according to the customer stratification classification levels, the merged results of the corresponding risk data, the merged results of the corresponding customer numbers, the merged results of the corresponding number percentages and the merged results of the corresponding overdue numbers, the customer stratification classification analysis results are constructed, and the mapping relationship between the customer stratification classification levels and the model scores of the two credit risk models is obtained. It should be noted that the preset risk data classification level rule is the customer stratification classification level.

[0047] In some embodiments, for example, taking Table 6 as an example, it can be seen from Table 6 that the highest risk value is 5.8% and the lowest risk value is 0.3%, which can be divided into 6 risk data classification levels, namely risk values ​​in [0, 1), [1, 2), [2, 3), [3, 4), [4, 5), [5, 6), and corresponding to level 1, level 2, level 3, level 4, level 5, and level 6 respectively. The grids with risk values ​​in Table 6 belonging to the same risk data classification level are merged to obtain the merged result of the risk values ​​as shown in Table 7.

[0048] Table 7 Merged results of risk values

[0049] Risk data classification Risk Value Level 6 5.80％ Level 5 4.45％ Level 4 3.35％ Level 3 2.39％ Level 2 1.45％ Level 1 0.50％ total 2.68％

[0050] Note: The calculation example of the number of customers is: Risk value for level 1 = (6+2) / (850+750)*100% = 0.5%.

[0051] In some embodiments, similarly, according to the 6 risk data classification levels, the grids within the customer number matrix, the number percentage matrix, and the overdue number matrix are merged to obtain the merged results of the corresponding customer number, the corresponding number percentage, and the corresponding overdue number as shown in Tables 8-10.

[0052] Table 8 Merged results of customer numbers

[0053] Risk data classification Number of customers Level 6 500 Level 5 5100 Level 4 4900 Level 3 5850 Level 2 7050 Level 1 1600 total 25000

[0054] Note: The calculation example of the number of customers is: the number of customers for level 1 = 850 + 750 = 1600 Table 9 Combined results of quantity proportion

[0055] Risk data classification Data share Level 6 2.00％ Level 5 20.40％ Level 4 19.60％ Level 3 23.40％ Level 2 28.20％ Level 1 6.40％ total 100.00％

[0056] Note: The calculation example of the quantity percentage is: quantity percentage for level 1 = 1600 / 25000*100% = 6.4%.

[0057] Table 10: Consolidated results of overdue amount

[0058] Risk data classification Overdue Level 6 29 Level 5 227 Level 4 164 Level 3 140 Level 2 102 Level 1 8 total 670

[0059] Note: An example of calculating the number of overdue payments is: the number of overdue payments for level 1 = 6 + 2 = 8.

[0060] In some embodiments, the customer stratification analysis results are constructed as shown in Table 11 based on the customer stratification level, the combined results of the corresponding risk data, the combined results of the corresponding number of customers, the combined results of the corresponding quantity ratio, and the combined results of the corresponding overdue number.

[0061] Table 11 Customer stratification analysis results

[0062]

[0063] In some embodiments, the mapping relationship between the customer stratification levels and the model scores of the two credit risk models is shown in Table 12.

[0064] Table 11 Mapping relationship between customer stratification levels and model scores of two credit risk models

[0065]

[0066]

[0067] It should be noted that the customer stratification analysis results need to satisfy the requirement that the risk data is monotonically sorted from low to high and has a stable distribution in different time periods. The stability can be seen from whether the customer distribution in different months is similar, or it can be judged by calculating the PSI value. For example, Table 12 shows the distribution of the customer stratification analysis results and risk data from January to May. It can be seen that the monthly distribution is basically the same, and the risk ranking of each month is monotonic, indicating that the customer stratification analysis results are stable and monotonic. If there is no good monotonic ranking in a certain month, this step can be merged within the matrix for adjustment, and the risk data classification level can be re-divided, or a third credit risk model and the obtained group dimension can be added to repeat the above method to obtain a customer stratification analysis result with good monotonic sorting and stable distribution.

[0068] Table 12 Distribution and risk data of customer stratification analysis results from January to May

[0069]

[0070]

[0071] Step S5: Based on the customer stratification analysis results, a customer structure analysis is performed on customers with no risk performance to obtain risk prediction results corresponding to customers with no risk performance.

[0072] In some embodiments, based on the customer stratification analysis results, a customer structure analysis is performed on customers with no risk performance to obtain risk prediction results corresponding to customers with no risk performance, including: scoring customers with no risk performance using two selected credit risk models to obtain two model scores corresponding to customers with no risk performance, and grading customers with no risk performance according to a mapping relationship between customer stratification grades and model scores of the two credit risk models to obtain a first customer stratification grade result; comparing the first customer stratification grade result with the customer stratification analysis results to obtain a first customer structure change identification result; based on the customer stratification analysis results, a risk estimation analysis is performed on new loans to customers with no risk performance, the product of the proportion of the number of customers at each customer stratification grade in the first customer stratification grade result and the risk value of the analysis sample data at the corresponding customer stratification grade is calculated, and the products obtained at all customer stratification grades are summed to obtain a risk prediction result corresponding to customers with no risk performance.

[0073] In some embodiments, the customer stratification and classification analysis results obtained above are applied to customers with no risk performance, and customer structure changes are identified based on the classification results. The specific steps are: (1) customers with no risk performance are scored using the two selected credit risk models to obtain two model scores corresponding to customers with no risk performance, and the level classification is completed based on the mapping relationship between the customer stratification level and the model scores of the two credit risk models. The classification results are shown in Table 13; (2) customer structure changes are identified. The customer distribution in Table 13 is basically consistent with the distribution of the analysis sample data, which shows that the customer structure is stable compared with the analysis sample data. It should be noted that there are two other situations in which the customer structure changes. The first situation is that the proportion of customers at the tail level (i.e., higher risk) increases, indicating that the quality of the customer base has declined. As shown in Table 14, compared with Table 11, the proportion of levels 4 to 6 in Table 14 has increased significantly; the second situation is that the proportion of customers at the head level (i.e., lower risk) increases, indicating that the quality of the customer base has improved. As shown in Table 15, compared with Table 11, the proportion of levels 1 to 3 in Table 15 has increased significantly.

[0074] Table 13 Results of customer stratification for customers without risk performance

[0075] Risk data classification Number of customers Data share Level 6 260 2.60％ Level 5 2000 20.00％ Level 4 1980 19.80％ Level 3 2350 23.50％ Level 2 2810 28.10％ Level 1 600 6.00％ total 10000 100.00％

[0076] Table 14 Data proportion of case 1

[0077] Risk data classification Data share Level 6 5.30％ Level 5 24.00％ Level 4 21.60％ Level 3 24.50％ Level 2 19.60％ Level 1 5.00％ total 100.00％

[0078] Table 15 Data proportion of situation 2

[0079] Risk data classification Data share Level 6 1.00％ Level 5 16.80％ Level 4 17.00％ Level 3 27.20％ Level 2 29.80％ Level 1 8.20％ total 100.00％

[0080] In some embodiments, the customer stratification analysis results are used to estimate the risk of new loans to customers with no risk performance. The specific steps are as follows: the customers with no risk performance have obtained the customer stratification level results, and the proportion of the number of customers in each customer stratification level in the customer stratification level results is multiplied by the risk value of the analysis sample data at the corresponding customer stratification level, and then the products obtained at each customer stratification level are added together to obtain the predicted risk results of customers with no risk performance. Table 13 shows the customer stratification level results of customers with no risk performance and the risk data of the analysis sample data at the same level. The predicted risk results of customers with no risk performance are calculated as follows: 5.8%*2.6%+4.45%*20%+3.35%*19.8%+2.39%*23.5%+1.45%*28.1%+0.5%*6%=2.7%.

[0081] Step S6: During the iterative impact assessment process of the risk strategy, based on the customer stratification analysis results, a customer structure analysis is performed on the placed customers to obtain risk prediction results corresponding to the placed customers.

[0082] In some embodiments, during the iterative impact assessment of the risk strategy, based on the customer stratification results, a customer structure analysis is performed on the placed customers to obtain risk prediction results corresponding to the placed customers, including: during the iterative impact assessment of the risk strategy, the placed customers are scored by using two selected credit risk models to obtain two model scores corresponding to the placed customers, and the placed customers are graded according to the mapping relationship between the customer stratification levels and the model scores of the two credit risk models to obtain a second customer stratification level result; based on the customer stratification analysis results, the product of the proportion of the number of customers at each customer stratification level in the second customer stratification level result and the risk value of the analysis sample data at the corresponding customer stratification level is calculated, and the products obtained at all customer stratification levels are summed to obtain the risk prediction results corresponding to the placed customers.

[0083] In some embodiments, when the risk control strategy is iteratively optimized, there will be a situation where the customer is replaced, which means that the customer refuses to be placed in the production strategy, but needs to be placed in the iterative optimization strategy. This type of customer has been rejected in the production and no loan has been issued, so the actual risk data cannot be obtained. Based on the customer stratification analysis results, the risk data of the analysis sample data of the corresponding customer stratification level can also be obtained according to the model scores of the two credit risk models. Therefore, the customer stratification level can be used to predict the risk of the strategically placed customers, and the risk estimation of such customers can be obtained. The specific steps are: score the placed customers through the two selected credit risk models, divide the placed customers according to the mapping relationship between the two model scores and the customer stratification level, multiply the proportion of the number of customers under each customer stratification level and the risk value of the analysis sample data under the corresponding customer stratification level, and then add the products obtained under each customer stratification level to obtain the predicted risk result of the placed customer group. The calculation method is the same as the calculation method in step S5, and will not be repeated here.

[0084] Step S6: Iteratively update the customer stratification analysis results.

[0085] In some embodiments, the customer stratification analysis results are iteratively updated, including: determining whether the credit risk model currently in use has risk identification; if the credit risk model currently in use does not have risk identification, then reselecting a credit risk model with a KS value greater than or equal to a preset threshold to replace the credit risk model without risk identification, and re-processing the analysis sample data for customer stratification level to obtain updated customer stratification analysis results; if the credit risk model currently in use has risk identification, but the risk identification effect is average, then adding a credit risk model with a KS value greater than or equal to the preset threshold, adjusting the preset risk data classification level rules, and re-merging the grids inside the original matrix to obtain updated customer stratification analysis results.

[0086] In some embodiments, with the passage of time and strategy iteration, the effect of the customer stratification analysis results will also decay, which is mainly manifested in the change of customer distribution in different time periods, or the monotonic sorting of risk data is not good. At this time, it is necessary to optimize the customer stratification level. Two situations are described here. The first is that the currently used credit risk model still has risk identification but the effect is poor or general (i.e., the slope of risk identification is examined). This situation refers to the risk data of the credit risk model having a monotonic sorting, but the risk differentiation is weak as shown in Table 16. As can be seen from Table 16, the weak risk differentiation is manifested as a slow slope of risk differentiation. At this time, a new credit risk model can be added to optimize the existing risk data classification rules (i.e., customer stratification level) using a matrix analysis method. The situation with good risk differentiation is shown in Table 17. As can be seen from Table 17, the good risk differentiation is manifested as a large slope of risk differentiation. The second is that the currently used credit risk model has no risk identification. This situation refers to the situation where the risk data is not monotonic. A new credit risk model can be selected to completely replace the existing credit risk model and re-process the customer stratification level.

[0087] Table 16: Case 1: Poor risk differentiation

[0088] Model points Risk Value 0-20 3.90％ 21-45 3.60％ 46-60 2.50％ 61-85 2.40％ 86-100 1.80％

[0089] Table 17 Case 1: Good risk differentiation

[0090] Model points Risk Value 0-20 4.50％ 21-45 3.60％ 46-60 2.50％ 61-85 1.80％ 86-100 1.20％

[0091] In summary, the present application provides a risk prediction method based on changes in customer structure. The method selects two credit risk models with KS values ​​greater than or equal to a preset threshold as the basis for customer stratification analysis and processing, and uses binning technology, matrix analysis technology and customer stratification analysis and processing technology to perform high-precision hierarchical classification of customer structure to obtain customer stratification analysis results, and applies the customer stratification analysis results to risk prediction for customers with no risk performance and placed customers, which can meet the needs of timeliness and accuracy, and continuously iterate and update the customer stratification analysis results. The updating method is simple and easy to implement, thereby providing effective technical support for risk monitoring in the field of consumer credit.

[0092] Embodiment 2

[0093] This embodiment provides a risk prediction system 200 based on customer structure changes based on the above embodiment 1. Figure 2 , a risk prediction method based on customer structure change of embodiment 1 is implemented, the system includes: a data acquisition module 210, used to obtain historical customers and their risk data, and construct analysis sample data; a binning processing module 220, used to select two credit risk models with KS values ​​greater than or equal to a preset threshold, and bin the analysis sample data according to the risk data through the selected two credit risk models, respectively, to obtain a first binning result and a second binning result; a matrix construction module 230, used to use the first binning result and the second binning result as rows and columns, respectively, to construct an original matrix, and calculate the customer quantity ratio and risk data of each grid in the original matrix; a matrix processing module 240, used to bin the analysis sample data according to the preset risk data According to the grading rules, the grids inside the original matrix are merged to obtain the merged results based on the group dimension, and the customer stratification analysis results are constructed, as well as the mapping relationship between the customer stratification levels and the model scores of the two credit risk models; the risk prediction module 250 is used to perform customer structure analysis on the placed customers based on the customer stratification analysis results during the iterative impact assessment of the risk strategy to obtain the risk prediction results of the corresponding placed customers, and to perform customer structure analysis on the placed customers based on the customer stratification analysis results during the iterative impact assessment of the risk strategy to obtain the risk prediction results of the corresponding placed customers; the update module 260 is used to iteratively update the customer stratification analysis results.

[0094] Embodiment 3

[0095] This embodiment further provides an electronic device based on the above embodiment 1. Figure 3 , Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0096] like Figure 3 As shown, the electronic device may include a processing device (e.g., a central processing unit, a graphics processor, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. In RAM 303, various programs and data required for the operation of the electronic device are also stored. The processing device 301, ROM 302, and RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0097] Typically, the following devices may be connected to the I / O interface 305: an input device 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, etc., an output device 307 including, for example, a liquid crystal display (LCD), a speaker, etc., a storage device 308 including, for example, a magnetic tape, a hard disk, etc., and a communication device 309. The communication device 309 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Figure 3 An electronic device having various devices is shown, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0098] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 309, or installed from the storage device 308, or installed from the ROM 302. When the computer program is executed by the processing device 301, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0099] Embodiment 4

[0100] Based on the above-mentioned embodiment 1, this embodiment further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned method are implemented.

[0101] It should be noted that the computer-readable medium in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0102] In this embodiment, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperTextTransferProtocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0103] The computer-readable medium may be included in the device, or may exist independently without being installed in the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: obtains training data, converts the training data into initial data; determines an initial rule base based on the initial data, and optimizes the parameters of the initial rule base to obtain a target rule base; calculates the rules in the target rule base according to a preset activation weight calculation formula to obtain activation weights; determines abnormal information according to the test data and the activation weights.

[0104] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0105] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0106] The units described in some embodiments of the present disclosure may be implemented by software or hardware. The units described may also be provided in a processor, for example, may be described as: a processor including a data acquisition unit, a rule determination unit, a weight calculation unit, and an anomaly determination unit. The names of these units do not, in some cases, constitute limitations on the units themselves, for example, the data acquisition unit may also be described as a "unit for acquiring training data".

[0107] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0108] The above contents are further detailed descriptions of the present invention in combination with specific implementation methods, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the scope of protection of the present invention.

[0109] Obviously, it should be understood by those skilled in the art that the above-mentioned various steps of the present invention can be performed in a manner different from the present invention, and the simulation method and experimental equipment include but are not limited to the above description. The above-mentioned various steps of the present invention can be performed in an order different from that here in some cases, and the steps shown or described above can be performed separately. Therefore, the present invention is not limited to any specific combination of hardware and software.

[0110] The above contents are further detailed descriptions of the present invention in combination with specific implementation methods, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the scope of protection of the present invention.

Claims

1. The risk prediction method based on changes in customer structure is characterized by: include: Obtain historical customer and risk data and construct analysis sample data; Select two credit risk models whose KS values ​​are greater than or equal to a preset threshold, and use the two selected credit risk models to bin the analysis sample data according to the risk data to obtain a first binning result and a second binning result; The first binning results and the second binning results are used as rows and columns respectively to construct the original matrix, and the customer quantity ratio and risk data of each grid in the original matrix are calculated; According to the preset risk data classification rules, the grids in the original matrix are merged to obtain the merged results based on the group dimension, and the customer stratification analysis results are constructed, and the mapping relationship between the customer stratification level and the model scores of the two credit risk models is obtained; Based on the customer stratification analysis results, conduct customer structure analysis on customers with no risk performance, and obtain risk prediction results for customers with no risk performance; During the iterative impact assessment process of risk strategies, based on the customer stratification analysis results, a customer structure analysis is conducted on the placed customers to obtain the risk prediction results for the corresponding placed customers.

2. The risk prediction method based on customer structure changes according to claim 1 is characterized in that: The two selected credit risk models are used to bin the analysis sample set according to the risk data to obtain a first binning result and a second binning result, including: The analysis sample set is binned according to the risk data through two credit risk models. The binning adopts the equal frequency binning method or the chi-square binning method to obtain the first binning result and the second binning result including model score, number of customers, proportion of quantity, number of overdue payments, and risk value.

3. The risk prediction method based on customer structure changes according to claim 2 is characterized in that: The first binning result and the second binning result are used as rows and columns respectively to construct the original matrix, and the customer quantity ratio and risk data of each grid in the original matrix are calculated, including: According to the four grouping dimensions of customer number, quantity proportion, overdue number and risk value, the model scores of the first binning results and the model scores of the second binning results are used as rows and columns respectively, and the customer number matrix, quantity proportion matrix, overdue number matrix and risk value matrix are constructed respectively. The customer quantity proportion of each grid in the quantity proportion matrix is ​​calculated, and the risk data of each grid in the risk value matrix is ​​calculated.

4. The risk prediction method based on customer structure changes according to claim 3 is characterized in that: The method of merging the grids in the original matrix according to the preset risk data classification rules to obtain the merging result based on the group dimension, constructing the customer stratification analysis result, and obtaining the mapping relationship between the customer stratification level and the model scores of the two credit risk models includes: Based on the preset risk data classification rules, the grids with similar risk values ​​in the risk value matrix are merged to obtain the merged result of the corresponding risk data; According to the mapping relationship between the grids inside the customer number matrix, the number ratio matrix, the overdue number matrix and the grids inside the risk value matrix, the grids inside the customer number matrix, the number ratio matrix and the overdue number matrix are merged respectively to obtain the merged results of the corresponding customer number, the merged results of the corresponding number ratio and the merged results of the corresponding overdue number respectively; Based on the mapping relationship between risk data classification levels and customer stratification classification levels, the customer stratification classification levels, the consolidated results of the corresponding risk data, the consolidated results of the corresponding number of customers, the consolidated results of the corresponding quantity ratios and the consolidated results of the corresponding overdue numbers are constructed to obtain the mapping relationship between the customer stratification classification levels and the model scores of the two credit risk models.

5. The risk prediction method based on customer structure changes according to claim 4 is characterized in that: Based on the customer stratification analysis results, the customer structure analysis is performed on the customers without risk performance to obtain the risk prediction results of the customers without risk performance, including: Using the two selected credit risk models, customers with no risk performance are scored to obtain two model scores corresponding to customers with no risk performance, and customers with no risk performance are graded according to the mapping relationship between the customer stratification grades and the model scores of the two credit risk models to obtain the first customer stratification grade result; Compare the first customer stratification level result with the customer stratification analysis result to obtain a corresponding first customer structure change identification result; Based on the results of customer stratification analysis, risk estimation analysis is conducted on new loans to customers with no risk performance. The product of the proportion of the number of customers in each customer stratification level in the first customer stratification level result and the risk value of the analysis sample data at the corresponding customer stratification level is calculated, and the products obtained at all customer stratification levels are summed to obtain the risk prediction results for customers with no risk performance.

6. The risk prediction method based on customer structure changes according to claim 4 is characterized in that: In the iterative impact assessment process of the risk strategy, based on the customer stratification results, a customer structure analysis is performed on the placed customers to obtain the risk prediction results corresponding to the placed customers, including: In the iterative impact assessment process of the risk strategy, the placed customers are scored by the two selected credit risk models, and the two model scores corresponding to the placed customers are obtained respectively. According to the mapping relationship between the customer stratification level and the model scores of the two credit risk models, the placed customers are graded to obtain the second customer stratification level result; Based on the customer stratification analysis results, calculate the product of the proportion of the number of customers at each customer stratification level in the second customer stratification level results and the risk value of the analysis sample data at the corresponding customer stratification level, and sum the products obtained at all customer stratification levels to obtain the risk prediction results for the corresponding placed customers.

7. The risk prediction method based on customer structure changes according to claim 1 is characterized in that: It also includes iterative updates of customer stratification analysis results, including: Determine whether the credit risk model currently in use has risk identification. If the credit risk model currently in use does not have risk identification, then reselect a credit risk model with a KS value greater than or equal to the preset threshold to replace the credit risk model without risk identification, and re-process the customer stratification and classification level of the analysis sample data to obtain updated customer stratification and classification analysis results; if the credit risk model currently in use has risk identification, but the risk identification effect is average, then add a credit risk model with a KS value greater than or equal to the preset threshold, adjust the preset risk data classification level rules, and re-merge the grids inside the original matrix to obtain updated customer stratification and classification analysis results.

8. A risk prediction system based on customer structure change that executes the risk prediction method based on customer structure change according to any one of claims 1 to 7, characterized in that: include: The data acquisition module is used to obtain historical customer and risk data and construct analysis sample data; A binning processing module is used to select two credit risk models whose KS values ​​are greater than or equal to a preset threshold, and bin the analysis sample data according to the risk data using the two selected credit risk models to obtain a first binning result and a second binning result; A matrix construction module, used to construct an original matrix using the first binning results and the second binning results as rows and columns respectively, and calculate the customer quantity ratio and risk data of each grid in the original matrix; The matrix processing module is used to merge the grids in the original matrix according to the preset risk data classification rules, obtain the merged results based on the group dimension, construct the customer stratification analysis results, and obtain the mapping relationship between the customer stratification levels and the model scores of the two credit risk models; A risk prediction module, which is used to perform a customer structure analysis on the placed customers based on the customer stratification analysis results in the iterative impact assessment process of the risk strategy, and obtain the risk prediction results of the corresponding placed customers; and to perform a customer structure analysis on the placed customers based on the customer stratification analysis results in the iterative impact assessment process of the risk strategy, and obtain the risk prediction results of the corresponding placed customers; The update module is used to iteratively update the customer stratification analysis results.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images