Credit rating prediction using a fuzzy MCDM approach with criteria interactions and TOPSIS sorting

Abstract

Multi-criteria decision making (MCDM) provides effective methods for dealing with the challenge of sorting credit ratings. This paper presents a novel data-driven MCDM sorting approach to predicting credit ratings. Our methodology combines the fuzzy TOPSIS-Sort-C model with the fuzzy best-worst approach, supported by a fuzzy cognitive map, to effectively deal with criteria interactions. This approach provides a corporate credit risk assessment, taking into account the uncertainties in credit risk assessment and relevance of its criteria by using fuzzy c-means and correlation-based feature selection. Our empirical analysis of 1138 US companies demonstrates the reliability of our model in dealing with a range of financial and non-financial indicators. The results demonstrate the potential of our methodology in credit rating assessment, with a good predictive performance relative to existing models.