| With the continuous development of Internet finance industry in recent years,P2 P online lending shows a growing trend.It is now popular not only among small and micro businesses,but also in other fields,such as medical cosmetology.With the rapid development of P2 P online loan market,its own problems have gradually emerged,the core of which is the risk assessment of whether the borrowers on the online loan platform will default.Once the borrower defaults,it will not only infringe on the interests of investors,but also have a great impact on the healthy development of online loan platforms.Therefore,the credit evaluation of borrowers in online loan platforms has obviously become the focus of attention of P2 P enterprises.This paper is based on the loan data of one of the major P2 P Lending platforms in the United States,Lending Club,for two years from 2017 to 2018.Data modeling is carried out through different algorithms,and the prediction effects of these models are compared.The advantages and disadvantages of different algorithms are obtained,so as to provide some method support for relevant platforms in China.This article first article 102960 sample data for pre-cleaning,unbalanced data dimension reduction,data processing and feature selection and so on a series of data preprocessing work,get 14 features,the second of these characteristics has a logistic regression,of random forests,XGBoost empirical study of the three single model,and then the parallel model based on three single models fusion of empirical research in the form of the comprehensive effect of the four models of their respective,effect of ranking score is calculated after the model,get the parallel fusion model of comprehensive empirical effect is better.Through empirical research,it is found that the default prediction effect is better by means of model fusion.It can contribute some ideas to the credit risk management of P2 P enterprises and has certain practical significance. |
PDF Full Text Request