Analysis of P2P Online Lending Default Based on Random Forest

Abstract

This article, based on Renren Loan website, will utilize SMOTE to process the debit and credit data in a balanced way. In addition, the importance for the variable-Random Forest and cross-validation conception will be applied for feature selection before the parameter optimization by grid searching, so that the model of basic Random Forest will be derived. Finally, featured modules of LDA will be added to further explore the reference value of loan description. The research findings show that the model has a satisfactory performance on test set in terms of predicting the results.