| As a new type of financial model,P2 P network lending breaks the limitations of traditional lending,such as high threshold,cumbersome procedures and small coverage.With ‘universal benefit' as core idea,it combines the Internet with private lending,so that both parties can match their borrowing and lending needs through third-party network platforms.This transaction not only alleviates the difficulty of financing and expensive financing for small and medium-sized enterprises in China,but also improves the quality and efficiency of financial services and promotes the construction of multi-level financial system in China.However,while promoting the realization of Inclusive Finance in China,P2 P network lending has gradually exposed some problems,especially the phenomenon of borrower default.Therefore,the analysis of P2 P network lending default risk,establishing a reasonable and effective default risk analysis model,and promoting the healthy and sustainable development of P2 P industry has become a research hotspot in the financial industry and academia.This paper first introduces the relevant theories of P2 P network lending default risk and the various methods used in the study of lending default risk.Secondly,it collects the historical lending transaction data published on Lending Club official website,and completes data preprocessing through a series of processes,such as data cleaning,data specification and feature coding to obtain dataset that can be used for modeling analysis.Then,the maximum information coefficient is introduced into the research of P2 P network lending default risk,and the complex network model based on the maximum information coefficient is constructed.By adjusting the threshold,observing the structural changes of the network,and analysing the possible influencing factors of borrower lending default from a global perspective.Finally,according to the maximum information coefficient between variables,the P2 P loan default prediction model based on different quantity of influencing factors is constructed by using machine learning frontier algorithm LightGBM,and the performance of the model is evaluated and compared.After a series of research and analysis,the following conclusions are obtained:(1)The maximum information coefficient,with its two important properties and advantages applicable to any type of variables,shows a good recognition ability in determining the important influencing factors of the variables under investigation.(2)In the complex network model of P2 P default risk,as the threshold increases,the important influencing factors of borrower default will be identified.The three most important factors are sub_grade,grade and int_rate.In addition,using P2 P default risk complex network model,we can also observe the changes of complex network structure and the connection between influencing factors.(3)By comparing the P2 P loan default prediction model based on different quantity of influencing factors,we found that the model with 39 influencing factors has relatively good performance.Taking the model with 39 quantitative factors as the final prediction model,it can not only identify the vast majority of performing borrowers,but also reduce default risk and default rate.Through sorting out and summarizing the research results of domestic and foreign scholars on P2 P network lending default risk,the research status and progress in this field are understood.On this basis,it develops ideas and explores new perspectives and new methods of problem analysis.There are two possible innovations in this paper:(1)Introducing the maximum information coefficient into the study of P2 P network lending default risk,which lays a foundation for the construction of default risk model.(2)Establishing a complex network model,considering all kinds of soft and hard information of borrowers and borrowing information in loan applications,and systematically studying and analyzing the possible influencing factors from a global perspective. |
PDF Full Text Request