Research On The Personal Credit Risk Assessment Of Online Peer-to-Peer Lending New Borrowers

Online peer-to-peer lending refers to the crediting behavior which is not play a mediating role of traditional financial institutions such as Banks, and then provide unsecured microfinance lending to the individual borrowers through the network platform directly. Online P2 P lending use the Internet platform to broaden scope, improve the efficiency of financing and to spread risks. Besides that, because of the low entry threshold, large range of floating interest rate and the duration of the flexible, P2 P lending meet the financing needs of individuals and SME.The research object of this paper is the new borrowers of network platform, which is affected by the information asymmetry easily. Credit risk assessment model of the new borrowers has more differences compared with the past research due to the new borrowers are lack of historical transaction records and repayment performance. In this paper, the concrete research content is as follows:(1) Explore the process and the characteristics of three basic business models, and mainly compares the basic preferences and risk control strategy of the domestic four representative platform business models, concludes their generalities and features. Summary the advantages and differences of discriminant analysis, Logistic regression, the linear programming method, neural network and decision tree method. Finally determine the LR regression as the borrower credit risk assessment method in this paper.(2) The research of credit risk assessment index system in the new borrower. Formed the borrower evaluate alternative indicators on the basis of the commercial bank personal credit risk assessment index system, and then fully considering the credit environment of the network characteristics and the content of new the borrower credit information. Through the "the ratio of good or bad customer " and ?2 statistic to grouping alternative indicators objectively.(3) The construction and inspection of credit risk assessment model in the new borrower. In order to reduce data operation, calculating the information gain to preliminary screening of alternative indicators, understanding the key feature vector's influence on the output variables. Next, in order to simplified model, it is given economic meaning for each group with WOE values, and replacing the virtual variables. Finally using the SPSS 17.0 to deal with sample data, establish Logistic regression model of new borrower credit risk assessment, then examine the GFI and prediction accuracy of the model.(4) The usage and improvement of credit risk assessment model in the new borrower. Select the real transaction records to presentation model in the actual business operation, and then put forward the improvement strategy according to the characteristics of defect model and the actual business model.