Support vector machines approach to pattern detection in bankruptcy prediction and its contingency

Abstract

This study investigates the effectiveness of support vector machines (SVM) approach in detecting the underlying data pattern for the corporate failure prediction tasks. Back-propagation neural network (BPN) has some limitations in that it needs a modeling art to find an appropriate structure and optimal solution and also large training set enough to search the weights of the network. SVM extracts the optimal solution with the small training set by capturing geometric characteristics of feature space without deriving weights of networks from the training data. In this study, we show the advantage of SVM approach over BPN to the problem of corporate bankruptcy prediction. SVM shows the highest level of accuracies and better generalization performance than BPN especially when the training set size is smaller. © Springer-Verlag Berlin Heidelberg 2004.