Research On Individual Credit Risk Assessment From The Perspective Of P2P Lending

P2P network credit platform refers to individuals and individuals achieve direct lending through the network,the main model is both sides of the borrowing and lending through the network for information and capital exchange.Because the lack of financing channels,the marketization of interest rate is relatively backward,individual investors have a huge market demand for financial management,it makes P2 P network loan in China's develop rapidly.However,due to regulatory problems and our personal credit mechanism is not perfect,making the P2 P platform bad debt rate is much higher than the traditional financial institutions,which requires us to do research and evaluation about it.Compared with the traditional bank loans,P2 P model have the following advantages: the lower threshold of access to both borrowing and lending;can set personalized trading contracts;yield significantly more than the bank financial products.But these reasons also objectively lead to the regulatory difficulties,so the platform's risk control system has become its core competitiveness.Firstly,this paper introduces the concept of credit and the main points of credit evaluation.Then it summarizes the origin and characteristics of the credit system in several countries,such as the United States,Europe and Japan,which have a relatively complete credit system in the world.The paper also introduces the current development trend of credit investigation in our country.Based on the analysis of the characteristics of P2 P credit system we establish a personal credit risk evaluation index system.We use three method to evaluate the personal credit risk,include support vector machine,GA-BP neural network and BP neural network.We also us 1000 groups of actual data to verify it.The results show that the accuracy of SVM is higher than that of neural network model in dealing with personal credit classification problem.The neural network model optimized by genetic algorithm also has higher accuracy and faster convergence speed than BP neural network model.In addition,in the binary classification problem,the accuracy of the three algorithms is high,but once extended to the three-class model,the accuracy will drop a lot,to find a suitable multiclassification algorithm has become the focus of future research.