Client Classification on Credit Risk Using Rough Set Theory and ACO-Based Support Vector Machine

Abstract

In the analysis of client classification based on support vector machine (SVM), redundant variables in the samples spoil the performance of the SVM classifier, two SVM parameters, C and sigma , must be carefully predetermined in establishing an efficient SVM model. This paper used rough sets as a preprocessor of SVM to select a subset of input variables and employed the ant colony optimization algorithm (ACO) to optimize the parameters of SVM. Additionally, the proposed ACO-SVM model that can automatically determine the optimal parameters was tested on the classification of the credit risk bank of listed companies in China. Then, we compared the accuracies of the proposed ACO-SVM model with those of other models of multivariate statistics and other artificial intelligence (BPN and fix-SVM). Experimental results showed that the ACO-SVM model performed the best classifying accuracy and generalization, implying that integrating the ACO with traditional SVM model is very successful.