Loan Default Prediction Using Diversified Sensitivity Undersampling

The loan default prediction is to predict rather the borrower will delay the repayment or not. This is an important problem for banking and finance companies. In this study, we focus on dealing with the data imbalance problem to enhance the performance of the loan default prediction. The approach in this study is a hybrid undersampling method that combines the clustering, the stochastic sensitivity measure and the radial basis function neural networks. A real loan default data from a P2P company in China is used to valid the performance of our method. Experiments results demonstrate that our approach yields a better generalization performance.