Author(s)

Yuqing Gu

Corresponding Author

Yuqing Gu

Abstract

Credit risk has always been one of the primary risks in the field of peer-to-peer (P2P) lending. As an attempt to address the problem, in this paper, we use the random forest method to conduct an empirical study on individual credit forecast using personal credit data from RenRenDai, a popular online P2P lending platform in China. Additionally, the five-fold cross-validation method is introduced to compare the effectiveness across different models including random forest, logistic regression and support vector machine (SVM). Results show that factors such as income, Marriage, birthplace, credit card limit, loan and credit card overdue, have significant effects on predicting personal credit risk; And the random forest model established is superior to logistic regression and support vector machine (SVM) in terms of Accuracy and Specificity.

Keywords

Credit Risk Management, Random Forest, Model Comparison, P2P Lending