Research On P2P Risk Control Model Based On Machine Learning Technology

Since the establishment of the first P2 P network lending platform in Zopa in the UK in 2005,the P2 P network lending platform has gained rapid development worldwide.Many P2 P network lending platforms have also been established,such as the United States Prosper,Kiva,Germany's Auxmoney,Japan's Aqush and so on.Lending Club successfully listed in the United States in December 2014,becoming the first listed P2 P network lending platform.Pat on the loan in 2007 marks the establishment of the P2 P network lending platform formally introduced China,and drive in the inclusive finance concept and national positive policies to stimulate the rapid development of micro credit market,become important innovation strength.P2P network lending is an important part of Internet banking,plays an important role in the financial disintermediation and finance to the center of.P2 P network lending platform since the introduction of our country,in the context of Inclusive Finance and services,small and micro enterprises to get rapid development,has become an important trend in the development of Internet banking.As a new type of micro financial institutions,P2 P network lending has become part of the multi-level capital market,enriching the traditional financial industry.As the financing model and financing channels of different,emerging Internet financial platform compared with the traditional financial model,with low barriers to entry,simple operation,investment risk prevention and control ability of poor.Its source of risk and dissemination also presents new features,P2 P,derived from Internet banking,is both a new Internet financial product and a means of risk prevention and control.In the country,including the current renrendai,ppdai are actively promoting the construction of data model for risk control,besides the control principle of small scattered by the risk control,is the core method of P2 P net loan risk control,analysis of different personal characteristics through the study data(i.e.data analysis)default rate corresponding to the nonlinear regression,through logic decision tree analysis,neural network modeling method to establish the control model and score card system data of wind,to grasp the different personal characteristics affect the corresponding default rate and the curing degree,to the approval of the decision engine and risk control in the business process,to guide the risk control approval to carry out the business,which is the goal and direction study.This paper uses the loan data of ppdai platform,through various machine learning algorithms to select the factors of the risk control model,and predicts the agreed default,and gets satisfactory results.Our model contains credit rating score replacing the credit level established by the company of ppdai,so we can verify our model.Although the recall rate is not high,it is compatible with the construction of the model.In the end,We draw the conclusion from the above analysis that the use of machine learning technology is able to correctly construct and optimize the current risk assessment model and further develop with the development of the microfinance industry and machine learning technology.Although the model is too simple,it can not be used in complicated scenes,but it is of great practical value as a research direction of the risk control model.