Credit risk evaluation using a weighted least squares SVM classifier with design of experiment for parameter selection

Support vector machines (SVM) is proved to be one of the most effective tool in credit risk evaluation. However, the performance of SVM is sensitive not only to the algorithm for solving the quadratic programming but also to the parameters setting in its learning machines as well as to the importance of different classes. In order to solve these issues, this paper proposes a weighted least squares support vector machine (LSSVM) classifier with design of experiment (DOE) for parameter selection for credit risk evaluation. In this approach, least squares algorithm is used to solve the quadratic programming, the DOE is used for parameter selection in SVM modelling and weights in LSSVM are used to emphasize the importance of difference classes. For illustration purpose, two publicly available credit datasets are selected to demonstrate the effectiveness and feasibility of the proposed weighted LSSVM classifier. The results show that the proposed weighted LSSVM classifier with DOE can produce the promising classification results in credit risk evaluation, relative to other classifiers listed in this study.