Using clustering-based bagging ensemble for credit scoring

Credit scoring has gained increasing attentions from banks, which can benefit from reducing possible risks of default. Many modeling techniques have been developed to improve the accuracy of credit scoring model. Based on the analysis of relationship between the performance of ensemble model and that of base classifiers, this paper presents a clustering-based ensemble model for credit scoring. The model uses clustering algorithm to enhance the diversity between the base classifiers, then choose base classifiers that meet the accuracy requirement to vote for the final decision. A real world credit dataset from UCI database is selected as the experimental data to demonstrate the accuracy of the model. The results show that clustering-based bagging ensemble model can significantly improved the efficiency in selection of base classifiers and generalization ability and thereby show enough attractive features for credit risk management system.