Research On Personal Credit Index Selection And Credit Risk Model Of P2P Network Lending

Since 2005,the P2P network lending model represented by Zopa has risen in Europe and the United States,and this model has been rapidly promoted worldwide.Compared with the P2P industry in Western countries,China's P2P industry disadvantage is that there is no perfect personal credit information system in the past.Most P2P industries use personal data collected by banks to make personal credit scores,which does not really reflect P2P.The characteristics of the industry,this paper takes a P2P data as the basis,and selects the index and credit risk assessment of the personal credit of the loan.First of all,this paper uses the "stepwise regression method" and "Botuta method" to combine the quantitative variable screening,and the qualitative variable is treated with "IV information value".Under the condition that the accuracy of the model can be guaranteed,the index that the customer needs to fill is reduced.The effect is to increase the satisfaction of users filling the form while reducing the operating costs of the platform.Secondly,the relevant indicator data obtained by the processing and screening of a P2P company borrower,including user basic attributes,occupation information,bank flow records,credit card billing behavior,loan information,user browsing behavior,and other types of information,through decision trees,BP The four machine learnings of neural network,BP-Adaboost and random forest empirically demonstrate the default of the borrower.The accuracy of the decision tree,random forest and BP-Adaboost are 80.51%,81.8%,85.68%and 89.01%,respectively.Finally,this paper combines the accuracy index,error rate,ROC curve and other evaluation indicators for analysis.The empirical results show that the model has obvious improvement after the screening,and BP-Adaboost has the best effect and accurately identifies 'good Bad' customer.This paper selects the evaluation indicators and evaluation methods that affect the credit of borrowers,and explains the importance of the indicators.This method can effectively identify‘good'customers and‘bad'customers,and provide a basis for P2P platform lending,which is beneficial to The P2P industry is better developed.