Research On P2P Credit Risk Assessment Model

Peer to peer lending refers to the borrowing before the individual and the individual.It realizes the issuance and lending of borrowing funds through the Internet.It is a direct transaction between the borrower and the lender without the intervention of traditional financial institutions.Therefore,it can be said that P2 P credit has created a new era of Internet finance,which makes up for the shortcomings of traditional financial institutions such as difficult and slow to make loans.It makes it easier for borrowers to obtain funds quickly and conveniently.At the same time,for lenders,it can obtain higher loan returns than traditional financial institutions such as banks.Therefore,it can be said that the birth of P2 P credit can better serve China’s real economy,and greatly promote the development of Internet finance in China.However,it is undeniable that China’s current online lending regulatory policy is not perfect and can not be relied on in many cases,resulting in many P2 P platforms using illegal fund-raising means to obtain illegal income,and this phenomenon is becoming increasingly serious,which has caused industry panic,and has seriously threatened the healthy and lasting development of the online lending market.。 Therefore,for the P2 P market,it is very urgent to construct the credit risk assessment of borrowers.Only by strengthening the credit assessment of borrowers,can the healthy development of the whole P2 P market be guaranteed.This paper studies and implements the construction of risk assessment model for P2 P borrowers.The main work is as follows:1.In view of the current mainstream modeling methods,this paper studies the applicability,advantages and disadvantages of each method,and finally chooses BP neural network to build the model.2.Combining the business model,wind control system,development situation and risk sources of P2 P credit in China,the credit evaluation index system of borrowers is formulated,and the evaluation index system is established according to each index and sample data.3.For the traditional BP neural network model based on gradient descent method,Levenberg-Marquardt(LM)algorithm is used to adjust and optimize the weights and thresholds,and the appropriate network parameters are determined to construct the credit risk assessment model.4.Two models are used to train the sample data,analyze and compare the experimental results,and select a better model to evaluate the borrower’s credit risk.Experiments show that Levenberg-Marquardt algorithm is superior to gradient descent method in the accuracy and convergence rate of borrower risk prediction,and the accuracy and convergence rate of prediction are high,which can be used in real business scenarios.