Research On Prediction Of Personal Credit Risk Of Internet Lending Based On Machine Learning

In 2007,P2 P online lending officially entered China.Over the past ten years,Internet finance has developed vigorously,and P2 P online lending has also emerged.In the early days of the development of P2 P online loan platforms,entrepreneurs were mostly Internet practitioners,and had no experience in lending and manipulation in the financial industry.Therefore,the lending of early P2 P platforms is mainly pure credit lending,but due to the imperfection of China’s credit system,the borrower’s default cost is very low,resulting in a high rate of borrower default.It is difficult for investors to trust.After 2012,the P2 P industry developed rapidly.As of the end of2013,the number of domestic P2 P platforms reached more than 800.The total transaction value of the entire P2 P lending industry exceeded 100 billion yuan,and P2 P entered a booming period of development.Although P2 P has developed rapidly afterwards,it is also mixed,and policies are frequently introduced at the regulatory level,but it has never landed.In 2016,frequent P2 P "thunderstorms" severely damaged the legitimate rights and interests of investors and caused a certain degree of negative impact on the sustainable and healthy development of the Internet financial industry.The P2 P platform,as the provider and collaborator of the information of the borrower and the lender,has alleviated the asymmetric information of the borrower and the lender to some extent.However,whether the disclosed borrower information can reflect the true credit status of the borrower still needs further research and discussion.Based on the loan data of Renrendai in 2014,this paper introduces the K-nearest neighbor algorithm,support vector machine,random forest,and neural network algorithm in machine learning to evaluate and predict the credit risk of platform borrowers.The empirical results show that machine learning algorithms can accurately assess the credit risk of borrowers in P2 P online lending,which means that in the future,machine learning algorithms can be used to quickly,cost-effectively and accurately assess the credit risk of small loans on small and medium-sized platforms.