Research On P2P Credit Default Classification Model Based On Genetic Algorithm

As a continuation and innovation in microfinance,P2P online lending has become an important part of inclusive finance and internet finance,supplementing the structure of traditional credit business.Therefore,since the beginning of 21^st century this industry has developed quickly in the world.According to industry maturity and supervision,the development of China's P2P online lending can be roughly divided into two stages.The first is during 2007-2015,P2P online lending is compatible with the background of China's development of inclusive finance and internet finance,while the supervision has lagged behind,so it has experienced rapid and barbaric growth.After 2015,China's supervision continued to exert its strength.At the same time,under the pressure of new economic and financial environment at home and abroad,this stage was characterized by the elimination of“inferior platform”,the loss of linkages and suspension of operations on a large number of platforms.As a result,the remaining platforms require better risk managements.So in the context of the improvement of supervision and the requirements of the industry itself to propose higher operational risk management capabilities,this paper will focusing on the research of the default prediction model about the platform users.This paper attempts to use the genetic algorithm with the capability of optimizing the feature engineering in the P2P online lending default prediction logistic model.First,this paper use the US Lending Club data to do the research,Through multiple genetic iterations to obtain the optimized model and it shows that the optimized model can significantly improve the accuracy,precision and recall rate.Then based on Chinese PPDai's data,comparing and analyzing the differences of US and China's finance environment,we find that many indicators of the model are also improved,which verifies the optimization.At last,this paper puts forward the following four suggestions for the development of China's P2P industry and the supervision:First,establish industry associations and open up P2P online lending platform data in a common use to prevent individual users from over-crediting.Second,make the lending data open source to absorb social computing power and research methods to enhance model accuracy.Third,we should innovate actively from the traditional credit analysis framework with the modern algorithms such as genetic algorithm in this new environment.New credit analysis indicators such as the circle of friends and the frequent location should be taken.Finally,promote legislation on data,Standardize Electronic Contract and Perfect Data Registration Mechanism.