A GA-based fuzzy decision tree approach for corporate bond rating

Abstract

The induction based on a tree structure is an appropriate representation of the complex human reasoning process such as a corporate bond rating application. Furthermore, the fuzzy decision tree (FDT) can handle the information about vague and incomplete classification knowledge represented in human linguistic terms. In addition, FDT is more flexible by relaxing the constraint of mutual exclusivity of cases in decision tree. We propose a hybrid approach using FDT and genetic algorithms (GA) enhances the effectiveness of FDT to the problem of corporate bond rating classification. This study utilizes a hybrid approach using GA in an attempt to find an optimal or near optimal hurdle values of membership function in FDT. The results show that the accuracy of the integrated approach proposed for this study increases overall classification accuracy rate significantly. We also show that the FDT approach increases the flexibility of the classification process. © Springer-Verlag Berlin Heidelberg 2004.