Research On The Prediction Of P2P Network Lending Probability

With the development of Internet security,the third party payment technology as well as people's increasing trust in online transactions,internet financial industry develops fast.P2 P network lending as a typical internet finance model,it provides financing channels for individuals and small enterprises,which rely on the third network platform.P2 P network lending provides such probability for people to carry out open and transparent micro credit transactions.It makes a useful contribution to the realization of financial disintermediation and the practice of inclusive finance.Since the rise of the market in 2005,it has been developing rapidly all over the world.But it also faces the problems of low success rate and high cost of capital,which hinder the further development of net loan market.Borrowing probability refers to borrowers can get enough lenders' attention and tender within the period specified by the third platform,and then raise the required funds.In this paper,we first give an analysis of the domestic and foreign literature,then we tries to explore the process of investors' investment decisions through the third net loan platform.We want to provide suggestion in the following areas,the first is how to improve the borrowing probability,the second is to help investors gain efficient investment.In the loan market,information dissemination and disclosure is essential to the development of financial institutions.Dealers do not know each other in online lending market.So the borrower usually has more advantages in the key information.They often have a large number of potential real information.Investors are in the inferior position of information.The information of both parties is asymmetric.The theory of asymmetric information is applied to the net loan industry.This paper starts from the network lending platform operation situation,examine the net loan business model,reproduce the P2 P net loan transaction process,to clarify the relationship between the borrower's information and the borrowing probability.This paper give a detailed introduction of domestic lending site-PPDai's borrowing mechanism,as the first pure credit unsecured net loan platform,PPDai is close to the nature of P2 P lending network.So we choose the platform as the research subject,and we select the data source of the empirical analysis with the network crawler software from this platform.Then paper selected indicators of 5 dimensions to perform cluster analysis on all the selected data in order to understand the use efficiency of the different characteristics of the borrower.The analysis results display that 8 credit rating of the borrower is divided into four categories.Then we know that people with the same credit rating may have different behavior pattern,and people who have different credit ratings may have similar behavior.Namely credit rating partly reflects the risk of default of the borrower.Therefore,it may be biased to use credit rating to define the dynamic behavior of borrowers.Therefore,to dig out the information which the signal of credit rating can't display and identify the performance of the user in the platform can help us find different patterns of behavior in the borrower as well as provides an opportunity to study at the probability of borrowing.We hope that through the study of the prediction model of the borrowing probability,we can use the open information of the borrower to identify the possibility of the borrower's borrowing.Then,the article selects the index to construct the prediction model of the full borrowing probability.We found that there is multicollinearity between the variables through the pretreatment of the sample data and the multicollinearity diagnosis..In order to solve the problem of multicollinearity among the explanatory variables,in order to make the final construction model more accurate,the paper uses the logical regression method of principal component selection after reading the relevant literature.,The core of this method is that the explanatory variables which may be correlated can be integrated into a few comprehensive indexes.Comprehensive indexes can reflect the original multivariate information.Finally we use the borrower's information as independent variables and the borrowing probability as the dependent variable to construct probability prediction model on the basis of the captured data.Through data processing,classification and quantification,we use principal component extracted instead of the original logistic to carry out regression.Then the extracted principal components are restored to study some other variables' influence direction and degree on the probability of full scale.And then we can provide the reference for the borrower to raise the probability of the full scale of the loan as well as to optimize the investors' decision.Finally,in view of the current low success rate of loans as well as the unsound credit mechanism,paper give some optimization countermeasures and future prospects.Including the following three aspects.We should improve the social credit system,strengthen the legislation in the Internet finance as well as improve the risk response mechanism in P2 P network lending.