Borrowers Credit Quality Rating Research Based On Extreme Learning Machines And FCM Algorithm

Credit quality rating is the comprehensive credit ability and moral quality factors to measure the borrower's credit status,so as to predict the borrower's default risk.In 2018,"thunderstorms" of Internet finance online loan platforms kept happening,one of the reasons was that a large number of borrowers defaulted,which seriously damaged the interests of investors and the platform and was not conducive to the healthy development of the online loan industry.Therefore,it is crucial to accurately measure the credit status of borrowers and control the default risk of borrowers.In this paper,from the construction of the borrower's credit quality rating index system,the construction of the borrower's credit quality score model to the borrower's credit quality rating logic,to reach the borrower's credit quality rating problem.The first is to use the evidence weight method and correlation coefficient method to screen the indicators and construct the credit quality rating index system.Secondly,borrower's credit quality score model based on the extreme learning machine constructed to calculate the borrower's credit quality score and predict the borrower's default risk.Three is according to the credit quality score and loss given default of the borrowers,FCM algorithm is used to cluster the borrowers with similar credit quality,and then the credit quality grade is divided to predict the borrower's default risk according to the double matching criterion of "the higher the credit quality grade,the lower the loss given default" and "the higher the credit quality grade,the lower the probability of default".This paper selects "renrendai" website of 7706 borrowers sample data for empirical research,the research results show that: Firstly,this paper constructs 14 credit quality rating indicators including "education background","loan amount","accumulative overdue amount","overdue times","income certification" and "loan certification" from four aspects of repayment ability,loan information,repayment willingness and credit certification;Secondly,the credit quality scoring model built based on extreme learning machine has a high ability of default identification.Second,the credit quality scoring model built based on ELM has a high ability of default identification.The overall accuracy of the model is 95.8%,and that of the defaulting customer is 88.3%.Thirdly,whether the credit quality grade is divided into 7grades,6 grades or 5 grades,this paper meets the dual matching criterion of "the higher the credit quality grade,the lower the loss given default" and "the higher the credit quality grade,the lower the probability of default ".Therefore,the credit quality grade classified in this paper can accurately evaluate the credit of the borrower and predict the default risk of the borrower.