SVDD: A proposal for automated credit rating prediction

Abstract

Credit rating prediction using clustering algorithms has become more and more important in the financial literature. Expanding the ideas of [4] and [5], we propose an approach to generate models for automated credit rating prediction based on support vector domain description (SVDD) and linear regression (LR). The models include the prediction for sovereign and corporate bonds. Another advantage is, the prediction models contain as many groups as rating grades exist, given by rating agencies like S&P, Fitch and Moody's. Our approach is formulated as a step-by-step procedure and all steps are illustrated by an example with artificial data. A numerical example with real data demonstrates the practical usability of our approach.