An Approach to Corporate Credit Rating Prediction Using Computational Intelligence-Based Methods

Abstract

Credit ratings tend to be very informative for investors and issuers and might serve as a powerful tool. The purpose of this paper is to investigate existing credit rating methodologies (e.g. Moody's, Standard and Poor's, Fitch) and to introduce improved data model for corporate ratings prediction based on computational intelligence methods. We hope that this study will provide academic researchers and industry practitioners new insights into the aspects of credit rating and its predictions. The research is performed on the selected companies that are constituents of the S&P 500 index. Company data from financial reports over period of 2016 to 2019 are analyzed and numerous financial indicators are included into analysis. The paper focuses on the design of data model, data preparation and working with missing values. Various well-known imputation techniques but also computational intelligence-based ones (e.g. fuzzy C-means) are applied to handle missing values and improve performance. In further research, the corporate credit rating prediction is brought down to a classification problem. Being a successful computational intelligence technique for credit ratings prediction, a typical neural network model is applied and compared to support vector machines as another popular data-based method in this domain. Finally, we have performed both cross-industry and industry-specific analysis. It is shown that industry-specific approach improved prediction results achieved by cross-industry data.