A rule adaptive dynamic adjustment method and system based on risk objectives

Abstract

This application relates to the technical field of credit management, and discloses a method and system for adaptively and dynamically adjusting rules based on risk targets. The method includes: S1. The early warning module comprehensively determines the credit risk value based on different key indicators. If the credit risk value is greater than the preset credit risk value threshold, proceed to the next step; S2. The division module collects different experience data; S3. The division module divides the different experience data into a first group and a second group, and a third group and a fourth group. The first model is trained and generated using the different experience data in the first group and the second group, and the second model is trained and generated using the different experience data in the third group and the fourth group; S4. The adjustment module determines the final risk result of the target customer by combining the output results of the first model and the second model, and sets a credit rule corresponding to the final risk result for the target customer. This application can dynamically adjust the credit rules of target customers and achieve better credit management effects.

Description

Technical Field

[0001] This application relates to the technical field of credit management, and particularly to a method and system for adaptively and dynamically adjusting rules based on risk targets. Background Art

[0002] With the development of computer application technology, intelligent credit management systems have become increasingly common. For example, the Chinese invention application with the publication number CN107924535A provides a credit management system, which includes: a unit for storing customer information according to customers; a storage unit for storing case information according to credit transactions, where the case information is associated with one or more customer information; a unit for receiving an identifier for determining a customer or a credit transaction; and a unit for notifying case information associated with a customer having the received identifier, or customer information associated with a credit transaction having the received identifier. In addition, the Chinese invention application with the publication number CN115358851A proposes a credit management method and system. The method includes: obtaining a credit instruction sent by a user and the user category to which the user belongs; selecting a system participating in credit management according to the user category to which the user belongs; and managing the credit instruction sent by the user based on the system participating in credit management. The technical solution improves the level of credit management and the ability to prevent credit risks, and also makes credit management more efficient and convenient.

[0003] However, neither of the above two invention applications can dynamically adjust the credit rules for target customers, which affects the effect of credit management. Summary of the Invention

[0004] When the credit risk value in this application is greater than a preset credit risk value threshold, different experience data is collected, and the different experience data is divided into a first group and a second group, as well as a third group and a fourth group. The different experience data in the first group and the second group are used to train and generate a first model, and the different experience data in the third group and the fourth group are used to train and generate a second model. The output results of the first model and the second model are combined to determine the final risk result of the target customer, and a credit rule corresponding to the final risk result is set for the target customer. This application aims to dynamically adjust the credit rules for target customers.

[0005] This application provides a method for adaptively and dynamically adjusting rules based on risk targets, which mainly includes the following steps:

[0006] S1. The early warning module collects and processes different key indicators of credit business. The different key indicators include the overall passing rate, the rule hit rate, the overdue rate, and the default rate. Based on the different key indicators, the credit risk value is comprehensively determined. When the credit risk value is greater than a preset credit risk value threshold, the next step is continued;

[0007] S2. The partitioning module collects different empirical data, which includes feature data generated from the customer's credit score, the customer's repayment history, the customer's income level, and the customer's loan purpose, as well as the corresponding risk results of the feature data. The risk results include the first risk, the second risk, and the third risk;

[0008] S3. The partitioning module divides different empirical data with corresponding risk results of the first risk and the second risk into the first group, divides different empirical data with corresponding risk results of the third risk into the second group, modifies the risk results of different empirical data in the first group to the first combined risk, and the partitioning module divides different empirical data with corresponding risk results of the first risk into the third group, divides different empirical data with corresponding risk results of the second risk and the third risk into the fourth group, modifies the risk results of different empirical data in the fourth group to the second combined risk. The partitioning module also uses different empirical data in the first group and the second group to train and generate the first model, and uses different empirical data in the third group and the fourth group to train and generate the second model;

[0009] S4. The adjustment module obtains the customer's credit score, the customer's repayment history, the customer's income level, and the customer's loan purpose of the target customer to generate feature data, and the adjustment module respectively inputs the feature data into the first model and the second model, and determines the final risk result of the target customer by combining the output results of the first model and the second model, and sets the credit rules corresponding to the final risk result for the target customer.

[0010] As a preferred technical solution of the present application, the partitioning module uses different empirical data in the first group and the second group to train and generate the first model, including the following steps:

[0011] S31. The partitioning module uses different empirical data in the first group and the second group to train and generate a first sub-model. When the first sub-model outputs +1, it represents that the input feature data is divided into the first combined risk. When the first sub-model outputs -1, it represents that the input feature data is divided into the third risk;

[0012] S32. For each empirical data in the first group and the second group, the partitioning module calculates the interval value between the empirical data and the partitioning surface of the first sub-model, and the partitioning module determines whether the interval values of different empirical data in the first group and the second group meet the preset conditions. If they meet, all steps are ended. If they do not meet, the next step is continued;

[0013] S33. The partitioning module partitions different empirical data with corresponding interval values greater than positive one into a first data group, partitions different empirical data with corresponding interval values greater than or equal to negative one and less than or equal to positive one into a second data group, partitions different empirical data with corresponding interval values less than negative one into a third data group, and the partitioning module sets the first data group, the second data group, and the third data group as target data groups;

[0014] S34. For each target data group, the partitioning module performs a first process on different empirical data in the target data group and determines whether the end flag is positive one. If so, all steps are ended. If not, this step is repeated.

[0015] As a preferred technical solution of the present application, the first process includes the following steps:

[0016] S341. The partitioning module uses different empirical data in the target data group to train and generate a new first sub-model, and for each empirical data in the target data group, the partitioning module calculates the interval value between the empirical data and the partitioning surface of the new first sub-model. The partitioning module also determines whether the interval values of different empirical data in the target data group meet a preset condition. If they meet, the end flag is set to positive one and all steps are ended. If they do not meet, the next step is continued;

[0017] S342. The partitioning module partitions different empirical data in the target data group with corresponding interval values greater than positive one into a new first data group, partitions different empirical data in the target data group with corresponding interval values greater than or equal to negative one and less than or equal to positive one into a new second data group, partitions different empirical data in the target data group with corresponding interval values less than negative one into a new third data group, and the partitioning module sets the new first data group, the new second data group, and the new third data group as target data groups.

[0018] As a preferred technical solution of the present application, in the process of the partitioning module determining whether the interval values of different empirical data in the first group and the second group meet the preset condition, the preset condition means that the number of interval values greater than or equal to negative one and less than or equal to positive one is zero.

[0019] As a preferred technical solution of the present application, the process of the adjustment module inputting the feature data of the target customer into the first model to obtain an output result includes the following steps:

[0020] S41. The adjustment module divides the feature data of the target customer using the first first sub-model, calculates the interval value between the feature data of the target customer and the division surface of the first first sub-model, and the adjustment module determines whether the current first sub-model is the last first sub-model. If so, proceed to S43. If not, determine the corresponding first sub-model according to the situation of the interval value corresponding to the feature data of the target customer;

[0021] S42. The adjustment module performs a second process on the feature data of the target customer, and the adjustment module determines whether the first sub-model determined by performing the second process is the last first sub-model. If so, proceed to the next step. If not, repeat this step;

[0022] S43. The adjustment module divides the feature data of the target customer using the last first sub-model, calculates the interval value between the feature data of the target customer and the division surface of the last first sub-model. If the interval value is greater than or equal to zero, divide the feature data of the target customer into the first combined risk. If the interval value is less than zero, divide the feature data of the target customer into the third risk.

[0023] As a preferred technical solution of the present application, the second process includes the following steps:

[0024] S421. The adjustment module determines whether the current first sub-model is the last first sub-model. If so, end all steps. If not, proceed to the next step;

[0025] S422. The adjustment module divides the feature data of the target customer using the current first sub-model, calculates the interval value between the feature data of the target customer and the division surface of the current first sub-model, and the adjustment module determines the corresponding first sub-model according to the situation of the interval value corresponding to the feature data of the target customer.

[0026] As a preferred technical solution of the present application, the process by which the adjustment module determines the final risk result of the target customer by combining the output results of the first model and the second model includes: when the output result of the first model is the first combined risk and the output result of the second model is the first risk, determining that the final risk result of the target customer is the first risk; when the output result of the first model is the first combined risk and the output result of the second model is the second combined risk, determining that the final risk result of the target customer is the second risk; when the output result of the first model is the third risk and the output result of the second model is the second combined risk, determining that the final risk result of the target customer is the third risk.

[0027] The present application also provides a rule adaptive dynamic adjustment system based on risk objectives, mainly including the following modules:

[0028] An early warning module, configured to collect and process different key indicators of credit business. The different key indicators include the overall passing rate, rule hit rate, overdue rate, and default rate, and comprehensively determine the credit risk value based on the different key indicators;

[0029] A partitioning module, configured to collect different experience data. The experience data includes characteristic data generated from the credit score of the customer, the repayment history of the customer, the income level of the customer, and the loan purpose of the customer, and the corresponding risk results of the characteristic data. The risk results include the first risk, the second risk, and the third risk, and is configured to partition different experience data with the corresponding risk results being the first risk and the second risk into the first group, partition different experience data with the corresponding risk results being the third risk into the second group, modify the risk results of different experience data in the first group to the first combined risk, partition different experience data with the corresponding risk results being the first risk into the third group, partition different experience data with the corresponding risk results being the second risk and the third risk into the fourth group, modify the risk results of different experience data in the fourth group to the second combined risk, and also use different experience data in the first group and the second group to train and generate a first model, and use different experience data in the third group and the fourth group to train and generate a second model;

[0030] An adjustment module, configured to obtain the credit score of the target customer, the repayment history of the target customer, the income level of the target customer, and the loan purpose of the target customer to generate characteristic data, respectively input the characteristic data into the first model and the second model, and determine the final risk result of the target customer in combination with the output results of the first model and the second model, and set a credit rule corresponding to the final risk result for the target customer.

[0031] The present application has at least the following beneficial effects:

[0032] In this application, first, the warning module collects and processes different key indicators of credit business, comprehensively determines the credit risk value based on different key indicators, and continues to the next step when the credit risk value is greater than the preset credit risk value threshold. Secondly, the division module collects different experience data. Thirdly, the division module divides different experience data with corresponding risk results of the first risk and the second risk into the first group, divides different experience data with corresponding risk results of the third risk into the second group, modifies the risk results of different experience data in the first group to the first combined risk, and the division module divides different experience data with corresponding risk results of the first risk into the third group, divides different experience data with corresponding risk results of the second risk and the third risk into the fourth group, modifies the risk results of different experience data in the fourth group to the second combined risk. The division module also uses different experience data in the first group and the second group to train and generate the first model, and uses different experience data in the third group and the fourth group to train and generate the second model. Finally, the adjustment module generates the characteristic data of the target customer, inputs the characteristic data into the first model and the second model respectively, determines the final risk result of the target customer by combining the output results of the first model and the second model, and sets the credit rules corresponding to the final risk result for the target customer. Through this application, not only can timely warnings be given when the credit risk value is large, but also the final risk result of the target customer can be automatically generated, and corresponding credit rules can be adopted for the target customer, so as to achieve better credit management effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of this application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0034] Figure 1 It is a flowchart of a rule adaptive dynamic adjustment method based on risk target in an embodiment of this application;

[0035] Figure 2 It is a schematic diagram of the first model of a rule adaptive dynamic adjustment method based on risk target in an embodiment of this application;

[0036] Figure 3 It is a schematic diagram of a rule adaptive dynamic adjustment system based on risk target in an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] The embodiments of the present application provide a method and system for adaptively and dynamically adjusting rules based on risk targets. The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and the above-mentioned drawings of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the term "including" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0038] It should be noted that all the acquisition and processing of information or data in this application are carried out on the premise of complying with the corresponding national data protection regulations and policies and obtaining the authorization given by the owner of the corresponding device.

[0039] For ease of understanding, please refer to Figure 1 , a method for adaptively and dynamically adjusting rules based on risk targets in the embodiments of the present application mainly includes the following steps:

[0040] S1. The early warning module collects and processes different key indicators of credit business. The different key indicators include the overall passing rate, rule hit rate, overdue rate, and default rate. Based on the different key indicators, the credit risk value is comprehensively determined. If the credit risk value is greater than the preset credit risk value threshold, proceed to the next step;

[0041] S2. The partitioning module collects different experience data. The experience data includes characteristic data generated from the customer's credit score, repayment history, income level, and loan purpose, as well as the risk results corresponding to the characteristic data. The risk results include the first risk, the second risk, and the third risk;

[0042] S3. The partitioning module partitions different empirical data with corresponding risk results of the first risk and the second risk into the first group, partitions different empirical data with corresponding risk results of the third risk into the second group, modifies the risk results of the different empirical data in the first group to the first combined risk, and the partitioning module partitions different empirical data with corresponding risk results of the first risk into the third group, partitions different empirical data with corresponding risk results of the second risk and the third risk into the fourth group, modifies the risk results of the different empirical data in the fourth group to the second combined risk. The partitioning module also uses the different empirical data in the first group and the second group to train and generate the first model, and uses the different empirical data in the third group and the fourth group to train and generate the second model;

[0043] S4. The adjustment module obtains the credit score of the target customer, the repayment history of the target customer, the income level of the target customer, and the loan purpose of the target customer to generate feature data. The adjustment module respectively inputs the feature data into the first model and the second model, and determines the final risk result of the target customer by combining the output results of the first model and the second model, and sets the credit rules corresponding to the final risk result for the target customer.

[0044] Specifically, in S1, the warning module collects different key indicators of credit business, including the overall passing rate, the rule hit rate, the overdue rate, and the default rate. Among them, the overall passing rate refers to the proportion of customers who are approved among the customers applying for credit within a certain period of time. The rule hit rate refers to the proportion of applications that meet specific approval rules during the credit review process. The overdue rate refers to the proportion of loans that are not repaid on time within a certain period of time in the total loans. The default rate refers to the proportion of customers who fail to fulfill their repayment obligations according to the contract terms within a certain period of time. The warning module comprehensively determines the credit risk value based on different key indicators. For example, the analytic hierarchy process can be used to determine the credit risk value. If the credit risk value is greater than the preset credit risk value threshold, proceed to S2. The credit risk value threshold is set according to the actual application scenario. In S2, the partitioning module collects different empirical data. The empirical data includes feature data and the risk results corresponding to the feature data. The feature data is generated based on the customer's credit score, the customer's repayment history, the customer's income level, and the customer's loan purpose and can be in the form of a vector. The risk results include the first risk, the second risk, and the third risk. It should be noted that the first risk corresponds to high risk, the second risk corresponds to medium risk, and the third risk corresponds to low risk. In S3, the partitioning module divides different empirical data with the corresponding risk results of the first risk and the second risk into the first group, divides different empirical data with the corresponding risk result of the third risk into the second group, modifies the risk results of different empirical data in the first group to the first combined risk, and keeps the risk results of different empirical data in the second group unchanged. In addition, the partitioning module divides different empirical data with the corresponding risk result of the first risk into the third group, divides different empirical data with the corresponding risk results of the second risk and the third risk into the fourth group, modifies the risk results of different empirical data in the fourth group to the second combined risk, and keeps the risk results of different empirical data in the third group unchanged. The partitioning module also uses different empirical data in the first group and the second group to train and generate the first model, uses different empirical data in the third group and the fourth group to train and generate the second model. How to obtain the first model will be introduced below. Based on the process of obtaining the first model, it is easy to think of the process of obtaining the second model. Therefore, the process of obtaining the second model will not be elaborated further below.In S4, the adjustment module obtains the credit score of the target customer, the repayment history of the target customer, the income level of the target customer, and the purpose of the loan of the target customer, aiming to generate feature data. The target customer can be any customer to be segmented. The adjustment module inputs the feature data into the first model and the second model respectively, and determines the final risk result of the target customer by combining the output results of the first model and the second model. The specific process will be described below. A credit rule corresponding to the final risk result is set for the target customer. For the sake of easy understanding, for example, if the final risk result is the first risk, credit rules such as rejecting or restricting loan applications and increasing interest rates can be adopted for the target customer.

[0045] Through the above method, not only can timely warnings be given when the credit risk value is large, but also the final risk result of the target customer can be automatically generated, and corresponding credit rules can be adopted for the target customer, so as to achieve better credit management effects.

[0046] Furthermore, the segmentation module uses different empirical data in the first group and the second group to train and generate the first model, including the following steps:

[0047] S31. The segmentation module uses different empirical data in the first group and the second group to train and generate a first sub-model. When the first sub-model outputs +1, it means that the input feature data is segmented into the first combined risk. When the first sub-model outputs -1, it means that the input feature data is segmented into the third risk.

[0048] S32. For each empirical data in the first group and the second group, the segmentation module calculates the interval value between the empirical data and the segmentation plane of the first sub-model, and the segmentation module judges whether the interval values of different empirical data in the first group and the second group meet the preset conditions. If they meet, all steps are ended. If they do not meet, the next step is continued.

[0049] S33. The segmentation module divides the different empirical data with corresponding interval values greater than +1 into the first data group, divides the different empirical data with corresponding interval values greater than or equal to -1 and less than or equal to +1 into the second data group, divides the different empirical data with corresponding interval values less than -1 into the third data group, and the segmentation module sets the first data group, the second data group, and the third data group as the target data groups.

[0050] S34. For each target data group, the segmentation module performs the first processing on the different empirical data in the target data group, and judges whether the end flag is +1. If it is, all steps are ended. If not, this step is repeated.

[0051] Specifically, in S31, the partitioning module uses different empirical data in the first group and the second group to train and generate a first sub-model. The first sub-model can be a support vector machine using soft margins. When the first sub-model outputs +1, it represents that the input feature data is partitioned into the first combined risk. When the first sub-model outputs -1, it represents that the input feature data is partitioned into the third risk. In S32, for each empirical data in the first group and the second group, the partitioning module calculates the interval value between the empirical data and the partitioning surface of the first sub-model. It should be noted that the interval value is also normalized so that the mean of the interval values corresponding to all the feature data partitioned into the first combined risk is +1, and at the same time, the mean of the interval values corresponding to all the feature data partitioned into the third risk is -1. The partitioning module determines whether the interval values of the different empirical data in the first group and the second group meet the preset conditions. The preset conditions will be introduced below. If they are met, all steps are ended. If not, a new first sub-model needs to be generated continuously, and S33 is continued. In S33, the partitioning module partitions the different empirical data with corresponding interval values greater than +1 into the first data group, partitions the different empirical data with corresponding interval values greater than or equal to -1 and less than or equal to +1 into the second data group, and partitions the different empirical data with corresponding interval values less than -1 into the third data group. And the partitioning module sets the first data group, the second data group, and the third data group as the target data groups. In S34, for each target data group, the partitioning module performs a first process on the different empirical data in the target data group. The process of the first process will be described below. Subsequently, it is determined whether the end flag is +1. If so, all steps are ended. If not, this step is repeated.

[0052] Further, the first process includes the following steps:

[0053] S341. The partitioning module uses different empirical data in the target data group to train and generate a new first sub-model. And for each empirical data in the target data group, the partitioning module calculates the interval value between the empirical data and the partitioning surface of the new first sub-model. The partitioning module also determines whether the interval values of the different empirical data in the target data group meet the preset conditions. In the case of meeting the conditions, the end flag is set to +1 and all steps are ended. In the case of not meeting the conditions, the next step is continued;

[0054] In S342, the partitioning module divides different empirical data in the target data group with a corresponding interval value greater than positive one into a new first data group, divides different empirical data in the target data group with a corresponding interval value greater than or equal to negative one and less than or equal to positive one into a new second data group, divides different empirical data in the target data group with a corresponding interval value less than negative one into a new third data group, and the partitioning module sets the new first data group, the new second data group, and the new third data group as the target data group.

[0055] Specifically, in S341, the partitioning module uses different empirical data in the target data group to train and generate a new first sub-model. The new first sub-model can also be a support vector machine with soft margin. For each empirical data in the target data group, the partitioning module calculates the interval value between the empirical data and the partitioning surface of the new first sub-model. The partitioning module also determines whether the interval values of different empirical data in the target data group meet the preset conditions. The preset conditions will be introduced below. If they are met, the end flag is set to positive one and all steps are ended. It should be noted that the initial value of the end flag is zero. If not, S342 continues. In S342, the partitioning module divides different empirical data in the target data group with a corresponding interval value greater than positive one into a new first data group, divides different empirical data in the target data group with a corresponding interval value greater than or equal to negative one and less than or equal to positive one into a new second data group, divides different empirical data in the target data group with a corresponding interval value less than negative one into a new third data group, and also sets the new first data group, the new second data group, and the new third data group as the target data group.

[0056] By executing the above method, the partitioning module can obtain a first model. In fact, the first model is composed of several first sub-models. For the sake of easy understanding, for example, the first model is as Figure 2 shown. A is the first first sub-model, B, D, E, F, G are all the last first sub-models, and C is the first sub-model.

[0057] Furthermore, in the process of the partitioning module determining whether the interval values of different empirical data in the first group and the second group meet the preset conditions, the preset condition means that the number of interval values greater than or equal to negative one and less than or equal to positive one is zero.

[0058] Specifically, in the process of the partitioning module determining whether the interval values of different empirical data in the first group and the second group meet the preset conditions, the preset condition means that the number of interval values greater than or equal to negative one and less than or equal to positive one is zero. It can also be added that the sign of the interval value of each empirical data is the same as the sign of its corresponding output result. In addition, the preset condition in the process of the partitioning module determining whether the interval values of different empirical data in the first group and the second group meet the preset conditions is the same as this.

[0059] Further, the process of the adjustment module inputting the characteristic data of the target customer into the first model to obtain an output result includes the following steps:

[0060] S41. The adjustment module uses the first first sub-model to divide the characteristic data of the target customer, and the adjustment module calculates the interval value between the characteristic data of the target customer and the division plane of the first first sub-model. The adjustment module also determines whether the current first sub-model is the last first sub-model. If so, continue with S43. If not, determine the corresponding first sub-model according to the situation of the interval value corresponding to the characteristic data of the target customer;

[0061] S42. The adjustment module performs a second processing on the characteristic data of the target customer, and the adjustment module determines whether the first sub-model determined by performing the second processing is the last first sub-model. If so, continue with the next step. If not, repeat this step;

[0062] S43. The adjustment module uses the last first sub-model to divide the characteristic data of the target customer, and calculates the interval value between the characteristic data of the target customer and the division plane of the last first sub-model. When the interval value is greater than or equal to zero, the characteristic data of the target customer is divided into the first combined risk. When the interval value is less than zero, the characteristic data of the target customer is divided into the third risk.

[0063] Specifically, for the convenience of understanding, continuing with the above example, in S41, the adjustment module uses the first first sub-model, that is, A, to divide the characteristic data of the target customer. The adjustment module calculates the interval value between the characteristic data of the target customer and the division plane of the first first sub-model, that is, A. The adjustment module also determines whether the current first sub-model, that is, A, is the last first sub-model. If so, continue with S43. If not, determine the corresponding first sub-model according to the situation of the interval value corresponding to the characteristic data of the target customer. Since A is not the last first sub-model, assume that the determined corresponding first sub-model is C. In S42, the adjustment module performs a second processing on the characteristic data of the target customer. Assume that the first sub-model determined by performing the second processing is E. The adjustment module determines whether the first sub-model determined by performing the second processing, that is, E, is the last first sub-model. If so, continue with the next step. If not, repeat this step. Since E is the last first sub-model, continue with the next step. In S43, the adjustment module uses the last first sub-model, that is, E, to divide the characteristic data of the target customer, calculates the interval value between the characteristic data of the target customer and the division plane of the last first sub-model. If the interval value is greater than or equal to zero, the characteristic data of the target customer is divided into the first combined risk. If the interval value is less than zero, the characteristic data of the target customer is divided into the third risk.

[0064] Further, the second process includes the following steps:

[0065] S421. The adjustment module determines whether the current first sub-model is the last first sub-model. If so, all steps are ended; if not, the next step is continued.

[0066] S422. The adjustment module uses the current first sub-model to divide the feature data of the target customer, calculates the interval value between the feature data of the target customer and the division surface of the current first sub-model, and the adjustment module determines the corresponding first sub-model according to the situation of the interval value corresponding to the feature data of the target customer.

[0067] Specifically, for the sake of easy understanding, the above example is continued. In S421, the adjustment module determines whether the current first sub-model, that is, C, is the last first sub-model. If so, all steps are ended; if not, the next step is continued. Since C is not the last first sub-model, the next step is continued. In S422, the adjustment module uses the current first sub-model, that is, C, to divide the feature data of the target customer, calculates the interval value between the feature data of the target customer and the division surface of the current first sub-model, and the adjustment module determines the corresponding first sub-model according to the situation of the interval value corresponding to the feature data of the target customer. Suppose the determined corresponding first sub-model is E.

[0068] Through the above method, the adjustment module can accurately divide the feature data of the target customer based on the first model, thus facilitating the adoption of credit rules for the final risk result of the target customer.

[0069] Further, the process of the adjustment module determining the final risk result of the target customer by combining the output results of the first model and the second model includes: when the output result of the first model is the first combined risk and the output result of the second model is the first risk, determining that the final risk result of the target customer is the first risk; when the output result of the first model is the first combined risk and the output result of the second model is the second combined risk, determining that the final risk result of the target customer is the second risk; when the output result of the first model is the third risk and the output result of the second model is the second combined risk, determining that the final risk result of the target customer is the third risk.

[0070] Specifically, the process of the adjustment module determining the final risk result of the target customer by combining the output results of the first model and the second model is introduced. If the output result of the first model is the first combined risk and the output result of the second model is the first risk, it is determined that the final risk result of the target customer is the first risk. If the output result of the first model is the first combined risk and the output result of the second model is the second combined risk, it is determined that the final risk result of the target customer is the second risk. If the output result of the first model is the third risk and the output result of the second model is the second combined risk, it is determined that the final risk result of the target customer is the third risk. At other times, it is considered that the final risk result of the target customer cannot be generated, and the first model and the second model can be regenerated.

[0071] The above describes a rule adaptive dynamic adjustment method based on risk objectives in the embodiments of the present application. Next, a rule adaptive dynamic adjustment system based on risk objectives in the embodiments of the present application will be described. Please refer to Figure 3 A rule adaptive dynamic adjustment system based on risk objectives in the embodiments of the present application includes the following modules:

[0072] The early warning module is used to collect and process different key indicators of credit business. Different key indicators include the overall passing rate, rule hit rate, overdue rate, and default rate, and comprehensively determine the credit risk value based on different key indicators;

[0073] The division module is used to collect different empirical data. The empirical data includes feature data generated from the customer's credit score, repayment history, income level, and loan purpose, as well as the corresponding risk results of the feature data. The risk results include the first risk, the second risk, and the third risk. And it is used to divide different empirical data with corresponding risk results of the first risk and the second risk into the first group, divide different empirical data with corresponding risk results of the third risk into the second group, modify the risk results of different empirical data in the first group to the first combined risk, divide different empirical data with corresponding risk results of the first risk into the third group, divide different empirical data with corresponding risk results of the second risk and the third risk into the fourth group, modify the risk results of different empirical data in the fourth group to the second combined risk, and also use different empirical data in the first group and the second group to train and generate the first model, and use different empirical data in the third group and the fourth group to train and generate the second model;

[0074] An adjustment module is used to obtain the credit score of a target customer, the repayment history of the target customer, the income level of the target customer, and the purpose of the loan of the target customer to generate feature data, input the feature data into the first model and the second model respectively, determine the final risk result of the target customer by combining the output results of the first model and the second model, and set a credit rule corresponding to the final risk result for the target customer.

[0075] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described system and modules can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0076] As described above, the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A rule adaptive dynamic adjustment method based on risk objectives, characterized in that: The steps include: S1. The early warning module collects and processes different key indicators of the credit business, including the overall pass rate, rule hit rate, overdue rate, and default rate. The credit risk value is comprehensively determined based on different key indicators. If the credit risk value is greater than the preset credit risk value threshold, the next step is continued; S2. The segmentation module collects different experience data, the experience data includes feature data generated by the customer's credit score, the customer's repayment history, the customer's income level, and the customer's loan purpose, and risk results corresponding to the feature data, the risk results include first risk, second risk, and third risk; S3, the division module divides different empirical data corresponding to the risk results of the first risk and the second risk into the first group, divides different empirical data corresponding to the risk results of the third risk into the second group, modifies the risk results of the different empirical data in the first group into the first combined risk, and the division module divides different empirical data corresponding to the risk result of the first risk into the third group, divides different empirical data corresponding to the risk results of the second risk and the third risk into the fourth group, modifies the risk results of the different empirical data in the fourth group into the second combined risk, and the division module further uses the different empirical data in the first group and the second group to train and generate the first model, and uses the different empirical data in the third group and the fourth group to train and generate the second model; S4. The adjustment module obtains the target customer's credit score, the target customer's repayment history, the target customer's income level, and the target customer's loan purpose to generate feature data, and the adjustment module inputs the feature data into the first model and the second model respectively, determines the target customer's final risk result based on the output results of the first model and the second model, and sets credit rules corresponding to the final risk result for the target customer.

2. The method for adaptive dynamic adjustment of rules based on risk targets according to claim 1, characterized in that: The division module uses different empirical data in the first group and the second group to train and generate a first model, including the following steps: S31, the classification module uses different experience data in the first group and the second group to train and generate a first sub-model, when the first sub-model outputs positive one, it represents that the input feature data is classified into the first combined risk, and when the first sub-model outputs negative one, it represents that the input feature data is classified into the third risk; S32, for each empirical data in the first group and the second group, the division module calculates the interval value between the empirical data and the division surface of the first sub-model, and the division module determines whether the interval values ​​of different empirical data in the first group and the second group meet the preset conditions, and if so, ends all steps, and if not, proceeds to the next step; S33, the division module divides different empirical data with corresponding interval values ​​greater than plus one into a first data group, divides different empirical data with corresponding interval values ​​greater than or equal to minus one and less than or equal to plus one into a second data group, and divides different empirical data with corresponding interval values ​​less than minus one into a third data group, and the division module sets the first data group, the second data group, and the third data group as target data groups; S34. Regarding each target data group, the division module performs a first processing on different experience data in the target data group, and determines whether the end flag is positive one. If yes, all steps are ended, and if not, this step is repeated.

3. The method for adaptive dynamic adjustment of rules based on risk targets according to claim 2, characterized in that: The first process includes the following steps: S341, the partitioning module uses different experience data in the target data group to train and generate a new first sub-model, and for each experience data in the target data group, the partitioning module calculates the interval value between the experience data and the partitioning surface of the new first sub-model, and the partitioning module also determines whether the interval value of different experience data in the target data group meets the preset condition, and if so, sets the end flag to positive one, and ends all steps, and if not, proceeds to the next step; S342, the division module divides the different empirical data in the target data group whose corresponding interval value is greater than plus one into a new first data group, divides the different empirical data in the target data group whose corresponding interval value is greater than or equal to negative one and less than or equal to positive one into a new second data group, divides the different empirical data in the target data group whose corresponding interval value is less than negative one into a new third data group, and the division module sets the new first data group, the new second data group, and the new third data group as target data groups.

4. The method for adaptive dynamic adjustment of rules based on risk targets according to claim 3, characterized in that: In the process where the division module determines whether the interval values ​​of different empirical data in the first group and the second group meet the preset condition, the preset condition refers to that the number of interval values ​​greater than or equal to negative one and less than or equal to positive one is zero.

5. The method for adaptive dynamic adjustment of rules based on risk targets according to claim 4, characterized in that: The process in which the adjustment module inputs the characteristic data of the target customer into the first model to obtain the output result includes the following steps: S41, the adjustment module uses the first first sub-model to divide the characteristic data of the target customer, calculates the interval value between the characteristic data of the target customer and the division surface of the first first sub-model, and the adjustment module determines whether the current first sub-model is the last first sub-model, and if so, proceeds to S43, and if not, determines the corresponding first sub-model according to the interval value corresponding to the characteristic data of the target customer; S42, the adjustment module performs a second process on the characteristic data of the target customer, and the adjustment module determines whether the first sub-model determined by the second process is the last first sub-model, and if yes, proceeds to the next step, and if no, repeats this step; S43, the adjustment module uses the last first sub-model to divide the characteristic data of the target customer, calculates the interval value between the characteristic data of the target customer and the division surface of the last first sub-model, and when the interval value is greater than or equal to zero, divides the characteristic data of the target customer into the first combination risk; when the interval value is less than zero, divides the characteristic data of the target customer into the third risk.

6. A rule adaptive dynamic adjustment method based on risk target according to claim 5, characterized in that: The second process includes the following steps: S421, the adjustment module determines whether the current first sub-model is the last first sub-model, and if so, ends all steps, and if not, proceeds to the next step; S422, the adjustment module uses the current first sub-model to divide the characteristic data of the target customer, calculates the interval value between the characteristic data of the target customer and the division surface of the current first sub-model, and the adjustment module determines the corresponding first sub-model according to the interval value corresponding to the characteristic data of the target customer.

7. The method for adaptive dynamic adjustment of rules based on risk targets according to claim 6, characterized in that: The process of the adjustment module determining the final risk result of the target customer in combination with the output results of the first model and the second model includes: when the output result of the first model is the first combined risk and the output result of the second model is the first risk, determining the final risk result of the target customer to be the first risk; when the output result of the first model is the first combined risk and the output result of the second model is the second combined risk, determining the final risk result of the target customer to be the second risk; when the output result of the first model is the third risk and the output result of the second model is the second combined risk, determining the final risk result of the target customer to be the third risk.

8. A rule-adaptive dynamic adjustment system based on risk targets, used to implement the method according to any one of claims 1 to 7, characterized in that: Includes the following modules: The early warning module is used to collect and process different key indicators of credit business, including overall pass rate, rule hit rate, overdue rate, and default rate, and comprehensively determine the credit risk value based on different key indicators; a partitioning module, for collecting different experience data, the experience data including feature data generated by a customer's credit score, a customer's repayment history, a customer's income level, and a customer's loan purpose, and risk results corresponding to the feature data, the risk results including a first risk, a second risk, and a third risk, and for partitioning different experience data corresponding to the risk results of the first risk and the second risk into a first group, partitioning different experience data corresponding to the risk results of the third risk into a second group, modifying the risk results of different experience data in the first group into a first combined risk, partitioning different experience data corresponding to the risk results of the first risk into a third group, partitioning different experience data corresponding to the risk results of the second risk and the third risk into a fourth group, modifying the risk results of different experience data in the fourth group into a second combined risk, and using different experience data in the first group and the second group to train and generate a first model, and using different experience data in the third group and the fourth group to train and generate a second model; The adjustment module is used to obtain the credit score of the target customer, the repayment history of the target customer, the income level of the target customer, and the loan purpose of the target customer to generate feature data, input the feature data into the first model and the second model respectively, and determine the final risk result of the target customer based on the output results of the first model and the second model, and set credit rules corresponding to the final risk result for the target customer.

Images