A peer-to-peer lending credit risk prediction method: Application of balanced relative margin machine

During the last few years, many classification methods have been applied to predicting peer-to-peer (P2P) lending credit risk. However, many models have given little consideration to data drifts and outliers. This paper examines the balanced relative margin machine (BRMM), and how it does in predicting default on a Chinese P2P lending platform, comparing its performance to the support vector machine (SVM) and logistic regression. In total, 55,596 P2P lending loans are examined. Discrimination performance for BRMM was found to be superior to SVM and logistic regression.