Research On The Matching Model Of The Two Parties In P2P Platform Considered On The Default Probability Of The Borrower

P2P network lending is a direct lending through the Internet platform.It is intended to reduce the financial threshold and allows more people to share financial services in the context of fair,just and open Internet.It will expand social financing,accelerate "financial disintermediation" and improve finance.The efficiency of resource allocation and the improvement of financial inclusiveness are all beneficial.However,due to the uncertainty and virtuality of the Internet environment,the credit risk of the P2P network lending platform is further amplified.Compared with the traditional lending model,the P2P network lending model faces greater credit problems and lending default risks.However,how to match borrowers and lenders on the basis of evaluating the credit risk of P2P platform borrowers has become the focus of P2P platform.Therefore,the focus of P2P online lending platforms is to control risks,screen out borrowers who are not easy to run away and have good credit conditions,and effectively matchs them with lenders,so as to contribute to the healthy development of the platform.It also helps to improve the matching efficiency of P2P online lending platforms.The matching model between lenders and borrowers in P2P online lending platforms is not completely consistent with the bank-enterprise matching relationship in commercial banks.At present,there are relatively few researches on the matching relationship between lenders and borrowers in P2P online lending platforms.Therefore,this thesis will focus on the characteristics of P2P platform both matching model,the thesis around the P2P lending platform borrower credit evaluation attribute characteristics,borrowers,lenders and borrowers default probability calculation model about the matching model of related research,to effectively guard against the risk of P2P platform,improve the matching efficiency of P2P platform is of great significance.In conclusion,this thesis mainly completed the following three research tasks:(1)It Established a Logistic Regression based credit attribute feature selection model to solve the problem of effectively discriminating borrowers' default status.The Logistic Regression model between the attribute and the default state was constructed,and the Wald statistics of the attribute characteristic regression coefficient were used to test the default discrimination ability of the attribute,and the attribute characteristics with weak default discrimination ability were deleted to ensure that all the attribute characteristics retained were those with strong default discrimination ability.(2)It used Random Forest-Logistic Regression to establish a borrower default probability measurement model to solve the problem of borrower default probability measurement.First of all,it established the borrower default discriminant model based on random forest model,through the analysis of the Decision Tree,Random Forests,Support Vector Machine,BP Neural Network and so on four kind of machine learning methods in the same sample under default discriminant ability,it selected the optimal model of Random Forest discriminant ability of default borrowers default discriminant model is set up;Secondly,considering that the random forest model could not determine the default probability of borrowers,a Logistic Regression model was established to calculate the default probability of P2P platform borrowers on the basis of the random forest default judgment results.(3)Based on the consideration of the default probability of the borrower,a matching model between the lender and the borrower was established through bilateral matching methods to solve the optimal matching problem between the lender and the borrower on the P2P platform.First of all,according to the difference between the actual value of each property characteristic of the borrower and the lowest acceptable value of the property characteristics of the borrower,the lender's satisfaction function for the borrower is constructed.The difference between the borrower's lowest acceptable value for each attribute characteristic of the lender,constructing the borrower's satisfaction function for the lender,and constructing the satisfaction function of the P2P platform according to the maximum return;Then,according to the three satisfaction functions,it structured both optimal matching model of objective function,the restrictions on the number of lenders a match object,it limited the number of borrowers a match object as constraint conditions,the optimal matching model to develop a lending both sides,to ensure the matching of P2P lending platform can meet lenders information requirements,it reduced the investment risk of lenders and improved the P2P lending platform;Finally,the validity of the optimal matching model between the borrower and the lender is verified by an example.The research in this paper laid a foundation for the research on the P2P platform's matching model of both parties,which took into account the default probability of borrowers.The specific application research reflected the guiding significance of this thesis for practice,and provided ideas and methods for expanding the research on other types of matching between the borrower and lender.