Credit evaluation method and device with good model interpretability, equipment and storage medium

Abstract

The invention provides a credit evaluation method and device with good model interpretability, equipment and a storage medium. The credit evaluation method comprises the following steps: constructinga credit evaluation model; training the credit evaluation model by using a training sample set with label information; evaluating a target sample by adopting the trained credit evaluation model to obtain an evaluation result; obtaining a plurality of neighbor samples of the target sample, and expanding each neighbor sample to form a neighbor sample set of the target sample; evaluating all samplesin the neighbor sample set by adopting the trained credit evaluation model to obtain a neighbor sample set with label information; and performing logistic regression on the neighbor sample set with the label information, and explaining an evaluation result of the credit evaluation model on the target sample according to a logistic regression result. According to the invention, the credit investigation condition of the target sample can be evaluated, and the evaluation result can be explained, so that the interpretability of the evaluation model is realized, and the evaluation model can be further improved and optimized.

Description

technical field [0001] The invention relates to the field of financial credit investigation, in particular to a credit evaluation method, device, equipment and storage medium with good model interpretability. Background technique [0002] The statements in this section merely mention background art related to the present disclosure and do not necessarily constitute prior art. [0003] In pan-financial fields such as credit, it is necessary to evaluate the credit of users. The traditional credit evaluation method requires a lot of manual (review specialist) participation, and has a high risk of privacy disclosure, human manipulation and fraud. For example, in the traditional credit evaluation method, the collection method of user data is mainly provided by the applicant himself, and then the approval agency manually verifies the accuracy of the information, and finally evaluates the applicant according to a set of evaluation methods established by itself. To carry out credi...

AI Technical Summary

Problems solved by technology

Although these credit evaluation models can perform more efficient and accurate evaluation of users' credit information, most of the evaluation models lack interpretability, that is, the evaluation models cannot fully and reasonably explain the evaluation results

For example, the current evaluation object (that is, the evaluation object corresponding to the target sample currently evaluated by the model) is evaluated as having credit risk, but the model cannot explain the reason for the evaluation result well, that is: which item or which of the evaluation object Several characteristics directly lead to the evaluation result

Images