Credit rating method

Abstract

The invention discloses a credit rating method which comprises the following steps: S1, acquiring a user credit information set, and extracting characteristic attributes of various pieces of information to form a characteristic attribute set; S2, performing multiple RIPPER classification on the characteristic attribute set, screening the characteristic attributes after RIPPER classification each time, and performing RIPPER classification again until the needed characteristic attribute set is obtained; S3, performing secondary screening on the characteristic attribute set, obtaining a final characteristic attribute set and performing RIPPER classification, and outputting a final RIPPER rating model; S4, inputting credit information of to-be-assessed users, extracting the characteristic attributes, inputting the characteristic attributes into the RIPPER rating model to be classified, thereby obtaining the credit rating result. According to the method disclosed by the invention, the RIPPER classification and characteristic screening twice are combined to establish the rating model, the realization method is simple, and an easy-to-understand rating rule can be conveniently acquired. Moreover, the processed data volume can be greatly reduced, and the rating efficiency and performance can be improved.

Description

technical field [0001] The invention relates to the technical field of credit evaluation, in particular to a credit evaluation method. Background technique [0002] Credit rating means that an independent third-party credit rating intermediary agency evaluates the ability and willingness of the debtor to repay the debt principal and interest in full on schedule, and uses simple rating symbols to indicate the severity of its default risk and loss, or to perform relevant contracts on the rated object. An overall assessment of the ability and willingness to commit and financial commitment. When the credit institution accepts the customer's credit application, it uses the characteristic variables in the application form submitted by the customer to establish a scoring model to obtain a credit value of the applicant, and compares this value with the standard value set in advance to judge the possibility of the borrower's overdue , so as to decide whether to grant credit and cred...

AI Technical Summary

Problems solved by technology

[0003] For users who newly submit applications, they need to perform credit ratings based on the relevant information provided. Currently, credit ratings for users mainly use credit scorecards or machine learning, but ratings based on scorecards are too general. However, the rating method based on machine learning is difficult to explain, and it is difficult for decision makers to intuitively understand the rules, which leads to difficult decision-making, and usually adopts empirical manual feature selection methods Or a simple feature selection algorithm. The size of the input data set of a classification task can be described by two parameters: the number of features N and the number of instances P. The analyzed data are often both N and P are large, and the large number of N and P will cause "Dimension disaster" and "combination explosion", the above feature selection methods for multi-dimensional feature attributes will lead to problems such as large task load, complex classification algorithm implementation, strong dependence and inflexibility, thereby reducing classification efficiency and not suitable for real-time performance In demanding credit rating applications

Images