Fuzzy Integrating Multiple SVM Classifiers and its Application in Credit Scoring

This paper presents a method of combining support vector machine (SVM) based on fuzzy integral. The classification has two steps: first map individual SVM classifiers´ decision values, which are good representatives of confidence, to memberships, second aggregate these memberships by fuzzy integral to obtain the final decision. Experimental results on two public datasets indicate that the performance of the proposed method outperforms the three conventional combining methods: single best, majority-rule ensemble and weighted-majority-rule ensemble. It clearly shows that the method has a great potential to find successful application in credit scoring area