A HYBRID APPROACH OF DEA, ROUGH SET THEORY AND RANDOM FORESTS FOR CREDIT RATING

In recent years, credit rating analysis has attracted lots of research interest in the literature. While the operating efficiency of a corporation is generally acknowledged to be a key contributor to the corporation's risk, it is usually excluded from early prediction models. To verify the operating efficiency as predictive variables, we propose a novel model to integrate rough set theory (RST) with the random forests (RF) technique, in order to increase credit rating prediction accuracy. In our proposed method, data envelopment analysis (DEA) is employed as a tool to evaluate the operating efficiency. Furthermore, the RST approach is used for variable selection due to its reliability in obtaining the significant independent variables, and utilized as a preprocessor to improve credit rating prediction capability by RF. The effectiveness of this methodology is verified by experiments comparing the RF, and compares the accuracy of the same prediction method with and without the DEA variable. The results show that operating efficiency does provide valuable information in credit rating predictions and the proposed approach provides better classification results.