Performance evaluation of least-squares probabilistic classifier for corporate credit rating classification problem

抄録

The corporate credit rating classification problem has attracted lots of research interests in the literature of financial risk management. This article introduces the least-squares probabilistic classifier to the problem in an attempt to provide a model with better explanatory power. Empirical results show that the least-squares probabilistic classifier outperforms the logistic regression model, random forest, and the support vector machine in prediction accuracy ratios and F1 scores, for the samples of bond issuer firms in Japan.