The User Feature Analysis And Overdue Forecast Of Network-lending Based On Data Mining

With the popularity and the widespread use of the Internet,Internet finance also developed rapidly accompanied by the continuous improving of Internet technology.From Yu’E Bao to peer-to-peer lending,network lending platform are springing up all over the place.However,due to China’s network credit industry development time is very short,and the personal credit rating system is not perfect and so on,making peer-to-peer lending platform is facing a great credit risk of overdue.How to control the overdue risk of borrowers is the key to the sustainable development of p2 p network borrowing platform.Therefore,based on massive transaction data and use the statistical methods and data mining techniques to analyse the characteristics of borrowers and establish an overdue forecasting model for a single borrower is a very important job for improving the operation of p2 p network borrowing platform.This paper mainly studies the application of statistical methods and data mining techniques in user characteristics analysis and overdue prediction of p2 p network borrowing platform.First of all,this article uses the Web crawler method to capture the real user data of a p2 p network platform,randomly selected 1319 full-scale borrowing data and uses the PAM clustering method partitions users and analyse user’s characteristics.Then,selects the features of overdue prediction model based on the user characteristics analysis and gain rate in engineering.Secondly,we use the Logistic regression,neural network model and support vector machine model to establish the forecasting model of the single borrower,and analyze the accuracy,stability and interpretability of the model.The result is that the neural network model has a good prediction accuracy but the model stability and interpretability are poor and the Logistic regression model has good stability but its accuracy is worse than the data mining method.In order to obtain high precision and stable overdue prediction model,establish neural network-Logistic regression model.Finally,summarize the borrower’s characteristics of China’s network lending industry and gives some advices.