Personal Credit Assessment Model Of Online Lending

Peer-to-peer lending(sometimes abbreviated P2 P lending)is a new financial practice in the Internet finance.Through lending a small amount of capital from investors to loan applicants on a third party platform,the practice can achieve mutual benefits for investors and investees.It has grown rapidly since 2007 when it entered China,and has become an essential part of Internet finance in China.Unlike bank credit,the new practice eases the issue that small enterprises and individuals are hard to get loan with no credit records.However,it also appears some defects.Especially it cannot effectually control credit risk,which leads to high overdue loan rates and bad debt rates.And some lending platforms went bankrupt due to a large quantity of unsecured loans.Therefore,there is an urge of controlling Internet finance to reduce P2 P lending risk.Adopting Big Data provides a new solution for P2 P lending.By Data Mining,it can dig out personal information like basic user information,family information and working situation.And it also can study personal behavior like consumption and social network habits.From the information about choosing the features that illustrate personal credit level to evaluate comprehensively can control P2 P lending risk and reduce losses from defaulting.The essay deals with the issues of P2 P lending platform risk control,by adopting Data Mining on personal credit assessment,as well as using Random Forest Algorithm to build personal credit assessment model,in order to forecast the tendency of defaulting,which can provide decision support for platforms.To ensure practical application value,the data in this essay comes from a large-sized P2 P lending Platform in China,using the latest three-month lending data as sample to empirical analyze.As the uncertainty of the data coming from the Internet,the original data has been pretreated,including standardizing continuous variable,quantizing attribute variables,interpolating missing values of samples.The original data has an severely imbalanced proportion of defaulting and non-defaulting samples.Therefore,the essay applies SMOTE Algorithm to balance the proportion of positive and negative samples.Then valuable indexes are obtained by Random Forest Model to pick up features with more clear classification effect,in order to improve the quality of Data Mining.While modeling,the optimized parameters of the model are chosen by Heuristic Algorithm,and the total model capability is assessed by Confusion Matrix.The optimal classification threshold are obtained by ROC(receiver operating characteristic)curve.At last,the defaulting probability,forecasted by Random Forest Algorithm,assesses loan applicant credit level for providing the P2 P lending platform with basis of decision to grant loans.