Asymmetric Random Subspace Method for Imbalanced Credit Risk Evaluation

In this paper, an asymmetric random subspace method is proposed for imbalanced credit risk evaluation. Asymmetric random subspace method integrates one of the advanced sampling methods, Synthetic Minority Over-sampling Technique (SMOTE), with random subspace to solve the imbalanced data problem. On the one hand, the advanced sampling method, SMOTE, combines informed over-sampling of the minority class with random under-sampling of the majority class and can solve the problems of basic sampling methods. On the other hand, it can also use random subspace strategy to enhance the performance of base classifiers. For illustration and verification purposes, two real world credit data sets are used as testing targets. Empirical results demonstrate that the proposed asymmetric random subspace method is a very promising method to the imbalanced credit risk evaluation.