A Dynamic Attribution Analysis Method for Credit Risk Based on Binary Tree

Abstract

The present invention provides a method for dynamically attributing credit risk based on a binary tree, belonging to the technical field of financial data. Specifically, it includes: determining the composition data of historical credit users corresponding to the surprise degree associated group during the index evaluation period, and combining the number of surprise degree associated groups and the change data of credit risk indicators during the index evaluation period to determine the explanation degree threshold of the surprise degree associated group. A binary tree is constructed based on the surprise degree associated group and the explanation degree threshold, and the generation of leaf nodes of the binary tree and the generation process of the attribution result of credit risk indicators are carried out based on the explanation degree and surprise degree of different surprise degree associated groups, improving the accuracy of attribution analysis.

Description

Technical Field

[0001] The present invention belongs to the technical field of financial data, and particularly relates to a method for dynamically attributing credit risk based on a binary tree. Background Art

[0002] In the field of credit risk control, the draw ratio, overdue rate, recovery rate, etc. are key indicators for measuring the operation level and risk control level of an enterprise. However, when these indicators show abnormal fluctuations, the enterprise needs to quickly locate which customer group causes this result. Traditional attribution analysis mainly relies on manual monitoring, analysis, and decision-making, which has problems such as low efficiency and untimely risk response.

[0003] In order to automatically achieve the attribution analysis of abnormal credit indicators, in the invention patent application CN202411974941.5 "A Method and System for Attribution Analysis of Changes in Overdue Indicators", the associated indicator characteristics are determined based on the changes in credit indicators, and the attribution analysis of the changes in credit indicators is carried out based on the sorting results of the correlation coefficients of the associated indicator characteristics, which improves the accuracy of the attribution analysis process. However, there are the following technical problems:

[0004] When conducting attribution analysis, due to the large number of credit indicators and the certain correlation between different credit indicators, if the method of drilling down dimensions is used to analyze and process the associated credit indicators, it will inevitably lead to the processing efficiency and difficulty of attribution analysis not meeting the requirements.

[0005] To solve the above technical problems, the present application provides a method for dynamically attributing credit risk based on a binary tree. Summary of the Invention

[0006] To achieve the purpose of the present invention, the present invention adopts the following technical solutions:

[0007] Specifically, the present application provides a method for dynamically attributing credit risk based on a binary tree, which specifically includes:

[0008] S1 Using the types of credit products, historical credit users are divided into multiple customer groups, and based on the changes in credit risk indicators in different unit time periods, the index stable period of the credit risk indicators is determined;

[0009] S2 Obtain the changes in the associated indicators of the customer groups in different index stable periods, and determine the surprise degree associated groups in the customer groups based on the changes in the associated indicators;

[0010] S3 determines the composition data of the historical credit users corresponding to the surprise degree associated group during the index evaluation period, and combines the number of the surprise degree associated group and the change data of the credit risk index during the index evaluation period to determine the interpretation degree threshold of the surprise degree associated group;

[0011] S4 constructs a binary tree based on the surprise degree associated group and the interpretation degree threshold, and generates the leaf nodes of the binary tree and the attribution result of the credit risk index based on the interpretation degree and surprise degree of different surprise degree associated groups.

[0012] The beneficial effects of the present invention are as follows:

[0013] Determining the surprise degree associated group in the customer group based on the change situation of the associated index during the stable index period realizes the screening of the customer group with frequent changes in the associated index, and further lays a foundation for excluding the customer group with frequent changes in the associated index during the attribution analysis to ensure the processing efficiency of the attribution analysis.

[0014] Generating the leaf nodes of the binary tree and the attribution result of the credit risk index based on the interpretation degree and surprise degree of different surprise degree associated groups fully considers the differences in the correlation degree between the change situation of the associated index and the change situation of the credit risk index for different fine-grained associated groups when using the binary tree for attribution analysis, and generates different binary tree leaf nodes, which improves the generation processing efficiency and accuracy of the binary tree leaf nodes, and also improves the processing efficiency of the attribution analysis.

[0015] A further technical solution is that the credit product type is determined according to the credit products developed by the credit service platform for different groups.

[0016] A further technical solution is that the historical credit users are divided into multiple customer groups, specifically including:

[0017] Customers belonging to the same credit product type are divided into the same customer group.

[0018] A further technical solution is that the credit risk indicators include the disbursement rate, the overdue rate, and the recovery rate.

[0019] A further technical solution is that the method for determining the index stable period of the credit risk index is as follows:

[0020] Taking the average value of the credit risk index on different dates in different unit periods to determine the benchmark value in different unit periods;

[0021] Determine the index change date in the date according to the deviation amount between the credit risk index on different dates in the unit time period and the reference value;

[0022] Determine whether the unit time period is the index stable period of the credit risk index according to the proportion of the number of the index change dates.

[0023] A further technical solution is that the index change date is the date when the deviation amount between the credit index and the reference value is not within the preset index deviation amount range.

[0024] A further technical solution is that when the proportion of the number of the index change dates in the unit time period is greater than the preset proportion of the number of change dates, it is determined that the unit time period does not belong to the index stable period of the credit risk index.

[0025] A further technical solution is to construct a binary tree, specifically including:

[0026] Obtain the associated index data, interpretation threshold, and customer group proportion threshold of the corresponding surprise degree associated group;

[0027] If the proportion of the number of historical credit users in the surprise degree associated group is less than the customer group proportion threshold, calculate the interpretation degree and surprise degree of the customer group, and use the customer group as the leaf node of the binary tree;

[0028] If the proportion of the number of historical credit users in the surprise degree associated group is not less than the customer group proportion threshold,

[0029] Calculate the interpretation degree and surprise degree of the associated index of the customer group, and when the interpretation degree of the surprise degree associated group does not meet the constraint condition of the interpretation threshold, use the surprise degree associated group as the leaf node of the binary tree;

[0030] When the interpretation degree of the surprise degree associated group meets the constraint condition of the interpretation threshold, based on the similarity degree of the basic information of the historical credit users of the subgroup of the surprise degree associated group, perform secondary division of the surprise degree associated group to obtain subgroups, construct leaf nodes of the binary tree based on the subgroups whose surprise degree meets the surprise degree threshold, and construct a decision tree with the leaf nodes of the binary tree.

[0031] A further technical solution is to divide historical credit users whose deviation amounts of basic information in a certain dimension are all within the preset deviation range into the same subgroup.

[0032] A further technical solution is that the basic information includes occupation, income, region, credit limit, draw data, and client.

[0033] A further technical solution lies in that the interpretability of the correlation index of the surprise correlation group is determined according to the ratio of the deviation of the correlation index of the surprise correlation group in the evaluation target period from the average value of the correlation index in the stable index period, and the deviation of the credit risk index in the evaluation target period from the average value of the correlation index in the stable index period.

[0034] A further technical solution lies in that the method for determining the surprise degree of the surprise correlation group is as follows:

[0035] Determine the benchmark value in different unit periods based on the average value of the credit risk indicators on different dates in different unit periods. Determine the index change date in the date according to the deviation of the credit risk indicator on different dates in the unit period from the benchmark value;

[0036] Determine the surprise degree of the customer group according to the ratio of the proportion of the number of index change dates in the evaluation target period to the proportion of the number of index change dates in different index stable periods.

[0037] A further technical solution lies in performing the generation process of the attribution result of the credit risk indicator, which specifically includes:

[0038] Take the customer group whose interpretability of the leaf node of the binary tree meets the interpretability threshold as the output result of the credit risk indicator.

[0039] Other features and advantages will be described in the subsequent specification. The objectives and other advantages of the present invention are achieved and obtained by the structure specifically pointed out in the specification and the drawings.

[0040] To make the above objectives, features, and advantages of the present invention more obvious and understandable, the following specific preferred embodiments are given, and in conjunction with the accompanying drawings, the detailed description is as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] By referring to the accompanying drawings and describing its exemplary embodiments in detail, the above and other features and advantages of the present invention will become more obvious.

[0042] Figure 1 is a flowchart of a method for dynamically attributing credit risk based on a binary tree;

[0043] Figure 2 is a flowchart of a method for determining the index stable period of the credit risk indicator;

[0044] Figure 3 is a flowchart of a method for determining the surprise correlation group in the customer group;

[0045] Figure 4It is a flowchart of a method for determining the explanatory threshold of the surprise degree associated population. Detailed implementation manners

[0046] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on the embodiments of this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this specification.

[0047] In the daily work process of a credit institution, once credit risk indicators such as the drawdown rate, overdue rate, and recovery rate change, due to the large number of customer groups, the credit institution often cannot accurately determine which customer groups' associated indicators have changed, resulting in the change of credit risk indicators, and thus cannot generate targeted intervention measures.

[0048] Specifically, if there is an abnormality in the Apple client, it will inevitably lead to a lower utilization rate of the loans of credit users with the Apple client, which may cause a change in the drawdown rate of the credit institution. Therefore, if the abnormal situation of the above customer group cannot be discovered in time and the client cannot be repaired in time, it is impossible to ensure that the drawdown rate of the credit structure remains within a reasonable range.

[0049] In this application, taking the surprise degree and the explanatory degree as variables, a binary tree model is used to screen customer groups with large changes in associated indicators, so that abnormal situations can be identified more quickly.

[0050] Taking the average value of credit risk indicators on different dates in different unit time periods to determine the benchmark value in different unit time periods, and determining the index change date in the date according to the deviation amount between the credit risk indicators on different dates in the unit time period and the benchmark value. The surprise degree of the customer group is determined according to the ratio of the proportion of the number of index change dates in the evaluation target time period to the proportion of the number of index change dates in different index stable time periods.

[0051] The surprise degree reflects the consistency between the daily changes of associated indicators and the changes in the indicator evaluation period.

[0052] The explanatory degree of the associated indicator is determined according to the ratio of the deviation amount between the average value of the associated indicators of the surprise degree associated population in the evaluation target time period and the average value of the associated indicators in the stable indicator time period, and the deviation amount between the credit risk indicator and the average value of the associated indicators in the evaluation target time period and the stable indicator time period, which reflects the change association between the associated indicator and the credit risk indicator.

[0053] Specifically, as Figure 1 shown, this application provides a dynamic attribution analysis method for credit risk based on a binary tree, which specifically includes:

[0054] S1 Use the type of credit product to divide historical credit users into multiple customer groups, and determine the stable period of the credit risk indicator based on the changes in the credit risk indicator in different unit time periods;

[0055] Determine the benchmark value in different unit time periods based on the average value of the credit risk indicator on different dates in different unit time periods. The indicator change date is the date when the deviation between the credit indicator and the benchmark value is not within the preset indicator deviation range. When the proportion of the number of indicator change dates in a unit time period is greater than 0.3, it is determined that the unit time period does not belong to the stable period of the credit risk indicator.

[0056] S2 Obtain the changes in the associated indicators of the customer groups in different stable periods of the indicators, and determine the surprise degree associated groups in the customer groups based on the changes in the associated indicators;

[0057] Determine the reference indicator data in the stable period of the indicator based on the average value of the associated indicators on different dates in the stable period of the indicator. When there is a stable period of the indicator whose deviation from the reference indicator data in other stable periods does not meet the requirements, it is determined that the customer group does not belong to the surprise degree associated group.

[0058] S3 Determine the composition data of the historical credit users corresponding to the surprise degree associated groups in the indicator evaluation period, and combine the number of the surprise degree associated groups and the change data of the credit risk indicator in the indicator evaluation period to determine the interpretation threshold of the surprise degree associated groups;

[0059] Determine the associated weight coefficients of different surprise degree associated groups according to the proportion of the number of historical credit users in different surprise degree associated groups, and use the surprise degree associated groups with the associated weight coefficients greater than the preset weight coefficient as the screened associated groups;

[0060] Determine the deviation between the credit risk indicator in the indicator evaluation period and the preset benchmark risk indicator threshold based on the change data of the credit risk indicator in the indicator evaluation period, determine the deviation rate based on the ratio of the deviation to the preset benchmark risk indicator threshold, determine the data matching coefficient based on the ratio of the number of the screened associated groups to the number of customer groups, determine the compensation factor based on the ratio of the deviation rate to the data matching coefficient, determine the compensation threshold based on the product of the preset interpretation threshold and the compensation factor, and determine the interpretation threshold of the surprise degree associated groups based on the difference between the compensation threshold and the preset interpretation threshold.

[0061] S4 constructs a binary tree based on the surprise degree associated groups and the interpretability threshold, and generates the leaf nodes of the binary tree and the attribution results of the credit risk indicators based on the interpretability and surprise degree of different surprise degree associated groups.

[0062] Furthermore, the credit product type is determined according to the credit products developed by the credit service platform for different groups.

[0063] Specifically, historical credit users are divided into multiple customer groups, specifically including:

[0064] Customers belonging to the same credit product type are divided into the same customer group.

[0065] It should be noted that the credit risk indicators include the disbursement rate, the overdue rate, and the recovery rate.

[0066] Specifically, as Figure 2 shown, the method for determining the stable period of the credit risk indicator is:

[0067] Determine the benchmark value in different unit periods based on the average value of the credit risk indicators on different dates in different unit periods;

[0068] Determine the date of indicator change in the date based on the deviation amount between the credit risk indicator on different dates in the unit period and the benchmark value;

[0069] Determine whether the unit period is the stable period of the credit risk indicator based on the proportion of the number of indicator change dates.

[0070] Furthermore, the indicator change date is the date when the deviation amount between the credit indicator and the benchmark value is not within the preset indicator deviation amount range.

[0071] It can be understood that when the proportion of the number of indicator change dates in the unit period is greater than the preset proportion of the number of change dates, it is determined that the unit period does not belong to the stable period of the credit risk indicator.

[0072] Specifically, as Figure 3 shown, the method for determining the surprise degree associated group in the customer group is:

[0073] Determine the associated indicators on different dates in the stable period of the indicator based on the change situation of the associated indicators of the customer group in the stable period of the indicator;

[0074] Determine the change date in the stable period of the indicator based on the change amount of the associated indicator with the adjacent date;

[0075] Determine whether the customer group is a surprise degree associated group based on the proportion of the number of change dates in different index stable periods.

[0076] Furthermore, the change date is a date when the change amount of the associated index with the adjacent date is greater than the preset index change amount threshold.

[0077] It should be noted that determining whether the customer group is a surprise degree associated group based on the proportion of the number of change dates in different index stable periods specifically includes:

[0078] Take the proportion of the number of change dates in different index stable periods as the associated index change coefficient in different index stable periods;

[0079] Take the average value of the associated index change coefficients in different index stable periods as the average change coefficient;

[0080] When the average change coefficient is greater than the preset change coefficient threshold, it is determined that the customer group does not belong to the surprise degree associated group.

[0081] Optionally, the method for determining the surprise degree associated group in the customer group is:

[0082] Based on the change situation of the associated index of the customer group in the index stable period, determine the associated index in different dates in the index stable period;

[0083] Determine the reference index data in the index stable period according to the average value of the associated indexes in different dates in the index stable period;

[0084] Determine whether the customer group is a surprise degree associated group based on the deviation amount of the reference index data between different index stable periods.

[0085] Furthermore, determining whether the customer group is a surprise degree associated group based on the deviation amount of the reference index data between different index stable periods specifically includes:

[0086] When there is an index stable period where the deviation amount of the reference index data from other index stable periods does not meet the requirements, it is determined that the customer group does not belong to the surprise degree associated group.

[0087] Optionally, the method for determining the surprise degree associated group in the customer group is:

[0088] S21 determines the associated metrics for different dates within the metric stable period based on the changes in the associated metrics of the customer group during the metric stable period. Based on the average value of the associated metrics for different dates within the metric stable period, it determines the reference metric data for the metric stable period. Using the deviation amount of the reference metric data between different metric stable periods, it determines the period change coefficient of the associated metrics of the customer group;

[0089] S22 determines the change dates within the metric stable period based on the change amount of the associated metrics with adjacent dates. Using the proportion of the number of change dates in different metric stable periods, it determines the associated metric change coefficient for different metric stable periods;

[0090] S23 determines the stability deviation coefficient of the associated metrics of the customer group based on the associated metric change coefficients and period change coefficients of different metric stable periods, and uses the stability deviation coefficient to determine whether the customer group is a surprise - associated group.

[0091] Furthermore, when the stability deviation coefficient does not meet the requirements, it is determined that the customer group does not belong to the surprise - associated group.

[0092] Optionally, the above step S21 includes the following content:

[0093] S211 determines the associated metrics for different dates within the metric stable period based on the changes in the associated metrics of the customer group during the metric stable period. Based on the average value of the associated metrics for different dates within the metric stable period, it determines the reference metric data for the metric stable period. When there is a metric stable period for which the deviation amount of the reference metric data from other metric stable periods does not meet the requirements, it is determined that the customer group does not belong to the surprise - associated group. When there is no metric stable period for which the deviation amount of the reference metric data from other metric stable periods does not meet the requirements, it proceeds to step S212;

[0094] S212 determines the data deviation amount of the reference metric data for different metric stable periods using the deviation amount of the reference metric data of the metric stable period from the reference metric data of other metric stable periods. When there is a metric stable period for which the data deviation amount does not meet the requirements, it proceeds to step S213. When there is no metric stable period for which the data deviation amount does not meet the requirements, it proceeds to step S214;

[0095] S213 When the number of index stable periods where the data deviation does not meet the requirements is greater than the preset number of stable periods, it is determined that the customer group does not belong to the surprise degree associated group. When the number of index stable periods where the data deviation does not meet the requirements is not greater than the preset number of stable periods, proceed to step S214;

[0096] S214 Based on the deviation of the reference index data between different index stable periods, determine the period change coefficient of the associated index of the customer group. When the period change coefficient of the associated index of the customer group does not meet the requirements, it is determined that the customer group does not belong to the surprise degree associated group. When the period change coefficient of the associated index of the customer group meets the requirements, proceed to step S22.

[0097] Optionally, the above step S22 includes the following content:

[0098] S221 Based on the change amount of the associated index with the adjacent date, when it is determined that there is no change date in different index stable periods, it is determined that the customer group belongs to the surprise degree associated group. When there is a change date in different index stable periods, proceed to step S222;

[0099] S222 Based on the proportion of the number of change dates in different index stable periods, determine the associated index change coefficient of different index stable periods. When the average value of the associated index change coefficients of different index stable periods does not meet the requirements, it is determined that the customer group belongs to the surprise degree associated group. When the average value of the associated index change coefficients of different index stable periods meets the requirements, proceed to step S223;

[0100] S223 When there is an index stable period where the associated index change coefficient does not meet the requirements, proceed to step S224. When there is no index stable period where the associated index change coefficient does not meet the requirements, proceed to step S225;

[0101] S224 When the number of index stable periods where the associated index change coefficient does not meet the requirements is greater than the preset number of stable periods, it is determined that the customer group does not belong to the surprise degree associated group. When the number of index stable periods where the associated index change coefficient does not meet the requirements is not greater than the preset number of stable periods, proceed to step S225;

[0102] S225 Based on the associated index change coefficients of different index stable periods and the time interval between different index stable periods, determine the index comprehensive change coefficient. When the index comprehensive change coefficient does not meet the requirements, it is determined that the customer group does not belong to the surprise degree associated group. When the index comprehensive change coefficient meets the requirements, proceed to step S23.

[0103] Further, the composition data of the historical credit users in the index evaluation period includes the number of historical credit users in different surprise correlation groups and the proportion of the number of historical credit users in the index evaluation period.

[0104] Specifically, as Figure 4 shown, the method for determining the interpretability threshold of the surprise correlation group is:

[0105] Based on the composition data of the historical credit users corresponding to the surprise correlation group in the index evaluation period, determine the proportion of the number of historical credit users in the index evaluation period;

[0106] Determine the correlation weight coefficients of different surprise correlation groups according to the proportion of the number of historical credit users in different surprise correlation groups, and use the surprise correlation groups with the correlation weight coefficients greater than the preset weight coefficient as the screened correlation groups;

[0107] Based on the change data of the credit risk index in the index evaluation period, determine the deviation amount between the credit risk index in the index evaluation period and the preset benchmark risk index threshold, determine the deviation rate based on the ratio of the deviation amount to the preset benchmark risk index threshold, and determine the interpretability threshold of the surprise correlation group according to the number of the screened correlation groups and the deviation rate.

[0108] Further, determining the interpretability threshold of the surprise correlation group according to the number of the screened correlation groups and the deviation rate specifically includes:

[0109] Determine the data matching coefficient based on the ratio of the number of the screened correlation groups to the number of the customer groups;

[0110] Determine the compensation factor according to the ratio of the deviation rate to the data matching coefficient, and determine the compensation threshold according to the product of the preset interpretability threshold and the compensation factor;

[0111] Based on the difference between the compensation threshold and the preset interpretability threshold, determine the interpretability threshold of the surprise correlation group.

[0112] Further, constructing a binary tree specifically includes:

[0113] Obtain the correlation index data, interpretability threshold, and customer group proportion threshold of the corresponding surprise correlation group;

[0114] If the proportion of the number of historical credit users in the surprise correlation group is less than the customer group proportion threshold, calculate the interpretability and surprise degree of this customer group, and use this customer group as the leaf node of the binary tree;

[0115] If the proportion of the number of historical credit users in the surprise correlation group is not less than the customer group proportion threshold,

[0116] Calculate the interpretability and surprise degree of the correlation indicators of this customer group, and when the interpretability of the surprise degree correlation group does not meet the constraint conditions of the interpretability threshold, use the surprise degree correlation group as the leaf node of the binary tree;

[0117] When the interpretability of the surprise degree correlation group meets the constraint conditions of the interpretability threshold, based on the similarity degree of the basic information of the historical credit users in the sub-group of the surprise degree correlation group, perform a secondary division of the surprise degree correlation group to obtain sub-groups, construct leaf nodes of the binary tree based on the sub-groups whose surprise degree meets the surprise degree threshold, and construct a decision tree with the leaf nodes of the binary tree.

[0118] Specifically, divide the historical credit users whose deviation amounts of basic information in a certain dimension are all within the preset deviation range into the same sub-group.

[0119] It should be noted that the basic information includes occupation, income, region, credit limit, utilization data, and client.

[0120] Furthermore, the interpretability of the correlation indicators of the surprise degree correlation group is determined according to the ratio of the deviation amount between the correlation indicators of the surprise degree correlation group in the evaluation target period and the average value of the correlation indicators in the stable indicator period, and the deviation amount between the credit risk indicators in the evaluation target period and the average value of the correlation indicators in the stable indicator period.

[0121] It should be noted that the method for determining the surprise degree of the surprise degree correlation group is as follows:

[0122] Determine the benchmark value in different unit periods based on the average value of the credit risk indicators on different dates in different unit periods, and determine the index change date in the date according to the deviation amount between the credit risk indicators on different dates in the unit period and the benchmark value;

[0123] Determine the surprise degree of the customer group according to the ratio of the proportion of the number of index change dates in the evaluation target period to the proportion of the number of index change dates in different index stable periods.

[0124] Furthermore, perform the generation process of the attribution result of the credit risk indicator, specifically including:

[0125] Use the customer group whose interpretability of the leaf node of the binary tree meets the interpretability threshold as the output result of the credit risk indicator.

[0126] Embodiment 1

[0127] In another possible embodiment, the method for determining the index stable period of the credit risk indicator is:

[0128] Determine the baseline value in different unit time periods based on the average value of credit risk indicators on different dates in different unit time periods;

[0129] Determine the deviation rate of different dates based on the ratio of the deviation amount between the credit risk indicator on different dates in the unit time period and the baseline value to the baseline value;

[0130] Determine whether the unit time period is the indicator stable period of the credit risk indicator according to the average value of the deviation rates on different dates in the unit time period.

[0131] Furthermore, when the average value of the deviation rates on different dates in the unit time period is greater than the preset deviation rate threshold, it is determined that the unit time period does not belong to the indicator stable period of the credit risk indicator.

[0132] Optionally, the method for determining the indicator stable period of the credit risk indicator is as follows:

[0133] S11 Determine the baseline value in different unit time periods based on the average value of credit risk indicators on different dates in different unit time periods, and use the deviation amount of the baseline value between the unit time period and other unit time periods to determine the baseline deviation coefficient of the unit time period;

[0134] S12 Determine the deviation coefficient of different dates based on the deviation amount between the credit risk indicator on different dates in the unit time period and the baseline value and the ratio to the baseline value, and combine the deviation amount between the credit risk indicators of different dates and adjacent dates, and determine the date change deviation coefficient of the unit time period based on the deviation coefficient of different dates;

[0135] S13 Determine the indicator change amount of the unit time period according to the mean value of the baseline deviation coefficient and the date change deviation coefficient of the unit time period, and use the indicator change amount to determine whether the unit time period is the indicator stable period of the credit risk indicator.

[0136] Furthermore, when the indicator change amount of the unit time period is greater than the preset change amount threshold, it is determined that the unit time period does not belong to the indicator stable period of the credit risk indicator.

[0137] Optionally, the following content is included in step S11 above:

[0138] S111 determines the baseline value in different unit time periods based on the average value of credit risk indicators on different dates in different unit time periods, and uses the deviation amount of the baseline value between the unit time period and other unit time periods to determine that when there are other unit time periods in which the deviation amount of the baseline value does not meet the requirements, it proceeds to step S112. When there are no other unit time periods in which the deviation amount of the baseline value does not meet the requirements in the unit time period, it proceeds to step S113;

[0139] S112 uses other unit time periods with a deviation rate from the baseline value of the unit time period that does not meet the requirements as deviation unit time periods. When the number of deviation unit time periods is greater than the preset number of deviation unit time periods, it determines that the unit time period does not belong to the index stable period of the credit risk indicator. When the number of deviation unit time periods is not greater than the preset number of deviation unit time periods, it proceeds to step S113;

[0140] S113 uses the deviation amount of the baseline value between the unit time period and other unit time periods to determine the baseline deviation coefficient of the unit time period. When the baseline deviation coefficient of the unit time period does not meet the requirements, it determines that the unit time period does not belong to the index stable period of the credit risk indicator. When the baseline deviation coefficient of the unit time period meets the requirements, it proceeds to step S12.

[0141] Optionally, the above step S12 includes the following content:

[0142] S121 determines the index change date in the unit time period based on the deviation amount between the credit risk indicators on different dates in the unit time period and the baseline value. When there is no index change date in the unit time period, it determines that the unit time period belongs to the index stable period of the credit risk indicator. When there is an index change date in the unit time period, it proceeds to step S122;

[0143] S122 determines that the unit time period does not belong to the index stable period of the credit risk indicator when the proportion of the number of index change dates in the unit time period is greater than the preset proportion of the number of change dates. When the proportion of the number of index change dates in the unit time period is not greater than the preset proportion of the number of change dates, it proceeds to step S123;

[0144] S123 determines the deviation coefficient of different dates based on the deviation amount between the credit risk indicators on different dates in the unit time period and the ratio to the baseline value, and combines the deviation amount between the credit risk indicators on different dates and adjacent dates. When the average value of the deviation coefficients of different dates does not meet the requirements, it determines that the unit time period does not belong to the index stable period of the credit risk indicator. When the average value of the deviation coefficients of different dates meets the requirements, it proceeds to step S124;

[0145] S124 determines the date change deviation coefficient of the unit time period based on the deviation coefficients of different dates. When the date change deviation coefficient is greater than the preset change deviation coefficient threshold, it is determined that the unit time period does not belong to the index stable period of the credit risk index. When the date change deviation coefficient is not greater than the preset change deviation coefficient threshold, it proceeds to step S13.

[0146] Specifically, the associated indicators of the customer group are determined according to the type of the credit risk index. Specifically, when the credit risk index is the overdue rate, the associated indicators are the overdue data of different customers in the customer group.

[0147] In Embodiment 2, optionally, the method for determining the interpretation degree threshold of the surprise degree associated group is as follows:

[0148] S31 determines the deviation amount between the credit risk index in the index evaluation period and the preset benchmark risk index threshold based on the change data of the credit risk index in the index evaluation period, and determines the deviation rate by the ratio of the deviation amount to the preset benchmark risk index threshold;

[0149] S32 determines the proportion of the number of historical credit users in the index evaluation period based on the composition data of the corresponding historical credit users of the surprise degree associated group in the index evaluation period, and determines the basic matching coefficient by using the proportion of the number of historical credit users.

[0150] S33 determines the associated weight coefficients of different surprise degree associated groups according to the proportion of the number of historical credit users in different surprise degree associated groups, and determines the data matching coefficient by the associated weight coefficients of different surprise degree associated groups and the proportion of the number of surprise degree associated groups in the customer group;

[0151] S34 determines the interpretation degree threshold of the surprise degree associated group according to the basic matching coefficient, the data matching coefficient, and the deviation rate.

[0152] Further, determining the interpretation degree threshold of the surprise degree associated group according to the basic matching coefficient, the data matching coefficient, and the deviation rate specifically includes:

[0153] Determine the matching coefficient by the sum of the basic matching times and the data matching coefficient, and use the ratio of the deviation rate to the matching coefficient as the deviation matching factor;

[0154] Use the preset interpretation degree threshold corresponding to the deviation matching factor as the interpretation degree threshold of the surprise degree group.

[0155] Optionally, before entering step S32, it is also necessary to determine whether the deviation rate is greater than a preset deviation rate threshold. If so, the interpretability threshold of the surprise group is determined using a set interpretability threshold. If not, step S32 is entered.

[0156] Optionally, the above step S32 includes the following:

[0157] S321 uses the composition data of the historical credit users corresponding to the surprise-related group during the index evaluation period to determine the proportion of the number of historical credit users in the index evaluation period, and uses the proportion of the number of historical credit users to determine the basic matching coefficient. When the basic matching coefficient is less than the preset matching coefficient threshold, the interpretability threshold of the surprise group is determined using the set interpretability threshold. When the basic matching coefficient is not less than the preset matching coefficient threshold, step S322 is entered;

[0158] S322 When the basic matching coefficient is within the preset matching coefficient range, step S323 is entered. When the basic matching coefficient is within the preset matching coefficient range, step S33 is entered;

[0159] S323 When the number of historical credit users corresponding to the surprise-related group during the index evaluation period is less than the preset number of credit users, the interpretability threshold of the surprise group is determined using the set interpretability threshold. When the number of historical credit users corresponding to the surprise-related group during the index evaluation period is not less than the preset number of credit users, step S33 is entered.

[0160] Optionally, the above step S33 includes the following:

[0161] S331 determines the association weight coefficients of different surprise-related groups according to the proportion of the number of historical credit users in different surprise-related groups. When the association weight coefficients of different surprise-related groups are all less than the preset weight coefficient, the interpretability threshold of the surprise group is determined using the set interpretability threshold. When there is a surprise-related group with an association weight coefficient not less than the preset weight coefficient, step S332 is entered;

[0162] S332 obtains the number of surprise-related groups with an association weight coefficient not less than the preset weight coefficient. When the number of surprise-related groups with an association weight coefficient not less than the preset weight coefficient is less than the preset number of association group threshold, step S333 is entered. When the number of surprise-related groups with an association weight coefficient not less than the preset weight coefficient is not less than the preset number of association group threshold, step S334 is entered;

[0163] S333 When the proportion of the number of the surprise-degree associated groups in the customer group is less than the preset group number proportion, determine the explanation degree threshold of the surprise-degree group by using the set explanation degree threshold. When the proportion of the number of the surprise-degree associated groups in the customer group is not less than the preset group number proportion, go to step S334;

[0164] S334 Determine the data matching coefficient based on the association weight coefficients of different surprise-degree associated groups and the proportion of the number of the surprise-degree associated groups in the customer group. When the data matching coefficient is less than the preset value of the matching coefficient, determine the explanation degree threshold of the surprise-degree group by using the set explanation degree threshold. When the data matching coefficient is not less than the preset value of the matching coefficient, go to step S34.

[0165] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the embodiments of the device, equipment, and non-volatile computer storage medium, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments.

[0166] The specific embodiments of this specification are described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be executed in a different order than in the embodiments and still achieve the desired results. Additionally, the processes depicted in the figures do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0167] The above is only one or more embodiments of this specification and is not used to limit this specification. For those skilled in the art, one or more embodiments of this specification can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of this specification shall be included within the scope of the claims of this specification.

Claims

1. A dynamic attribution analysis method for credit risk based on binary tree, characterized in that, Specifically, it includes: Dividing historical credit users into multiple customer groups according to the types of credit products, and determining the index stable period of the credit risk index based on the changes in the credit risk index in different unit time periods; Obtaining the changes in the associated indicators of the customer groups in different index stable periods, and determining the surprise degree associated groups in the customer groups based on the changes in the associated indicators; Determining the composition data of the historical credit users corresponding to the surprise degree associated groups in the index evaluation period, and combining the number of the surprise degree associated groups and the change data of the credit risk index in the index evaluation period to determine the interpretation degree threshold of the surprise degree associated groups; Constructing a binary tree based on the surprise degree associated groups and the interpretation degree threshold, and generating the leaf nodes of the binary tree and the attribution results of the credit risk index based on the interpretation degree and surprise degree of different surprise degree associated groups; The surprise degree reflects the consistency between the daily changes in the associated indicators and the changes in the index evaluation period; The interpretation degree reflects the change association between the associated indicators and the credit risk index; The associated indicators of the customer groups are determined according to the types of the credit risk index; Constructing a binary tree specifically includes: Obtaining the associated indicator data, interpretation degree threshold, and customer group proportion threshold of the corresponding surprise degree associated groups; If the proportion of the number of historical credit users in the surprise degree associated group is less than the customer group proportion threshold, calculating the interpretation degree and surprise degree of the customer group, and taking this customer group as the leaf node of the binary tree; If the proportion of the number of historical credit users in the surprise degree associated group is not less than the customer group proportion threshold, Calculating the interpretation degree and surprise degree of the associated indicators of the surprise degree associated group, and when the interpretation degree of the surprise degree associated group does not meet the constraint conditions of the interpretation degree threshold, taking the surprise degree associated group as the leaf node of the binary tree; When the interpretation degree of the surprise degree associated group meets the constraint conditions of the interpretation degree threshold, based on the similarity of the basic information of the historical credit users of the sub-groups of the surprise degree associated group, performing secondary division of the surprise degree associated group to obtain sub-groups, constructing the leaf nodes of the binary tree based on the sub-groups whose surprise degree meets the surprise degree threshold, and constructing a binary tree with the leaf nodes of the binary tree; Taking the customer groups whose interpretation degree of the leaf nodes of the binary tree meets the interpretation degree threshold as the output results of the credit risk index.

2. The dynamic attribution analysis method for credit risk based on binary tree according to claim 1, characterized in that The types of credit products are determined according to the credit products developed by the credit service platform for different groups.

3. The dynamic attribution analysis method for credit risk based on binary tree according to claim 1, characterized in that Dividing historical credit users into multiple customer groups specifically includes: Dividing customers belonging to the same type of credit product into the same customer group.

4. The dynamic attribution analysis method for credit risk based on binary tree according to claim 1, wherein The credit risk indicators include the disbursement rate, overdue rate, and recovery rate.

5. The dynamic attribution analysis method for credit risk based on binary tree according to claim 1, characterized in that The method for determining the index stable period of the credit risk index is: Determining the benchmark values in different unit time periods based on the average values of the credit risk indicators on different dates in different unit time periods; Determining the index change dates in these dates according to the deviation amounts between the credit risk indicators on different dates in the unit time period and the benchmark values; Determine whether the unit time period is an index stable period of the credit risk index according to the proportion of the number of the index change dates.

6. The method for dynamically attributing credit risk based on a binary tree according to claim 5, wherein The index change date is a date when the deviation amount between the credit index and the reference value is not within the preset index deviation amount range.

7. The dynamic attribution analysis method for credit risk based on a binary tree according to claim 1, characterized in that The basic information includes occupation, income, region, credit limit, drawdown data, and client side.

Images