Default Risk Prediction Of A P2P Platform Based On Machine Learning

At present,the Internet finance industry is booming.The lending business as a basic financial business.whose ideas and operating modes are constantly being improved in the financial industry.In the micro-finance market,the new concept of"inclusive financial" has been spread,which creates P2P network credit platform.In the meantime,the platform provides a very simplified process of loan review and the tracking records of the loan project,which provides a convenient lending channel for investors and borrowers.The platform shortens the processing time of traditional lending business.However,with the P2P industry changing from the growth phase to the explosion phase,the industry competition is increasingly becoming fierce.Unregulated industry operations generate many problematic platforms and high default rates,which make platform operators suffer huge losses.In order to prevent further deterioration of the development of the industry,besides the corresponding implementation of regulatory policies,it's necessary to establish and operate a cost-effective model of default risk prediction.The paper firstly introduces the development background and research status of the P2P industry,and then explains the relevant theoretical knowledge of the machine learning classification model.Secondly,I uses the Octopus Reptile Software to collect the open source loan data of the“Renren Dai”platform,and performs preprocessing work on the data.And,the paper carried out statistical analysis between the variable and the customer default rate.Then,I built three machine learning models,including ecision tree,random forest and XGBoost,and evaluated the risk prediction ability of each model.Based on the adjusted XGBoost model,which has high precision and high recall rate.And I combine with the strong aggregation of the Stacking integration model,which finally established a combined default risk prediction model based on the optimized XGBoost model.The results of empirical analysis show that the model can provide theoretical and technical support for the lending audit module in the P2P industry.It will promote the healthy and orderly development of P2P credit platform.