IVIFCM-TOPSIS for bank credit risk assessment

Abstract

Bank credit risk assessment is performed by credit rating agencies in order to reduce information asymmetry in financial markets. This costly process has been automated in earlier studies by using systems based on machine learning methods. However, such systems suffer from interpretability issues and do not utilize expert knowledge effectively. To overcome those problems, multi-criteria group decision-making (MCGDM) methods have recently been used to simulate the assessment process performed by the committee of multiple credit risk experts. However, standard MCGDM methods fail to consider high uncertainty inherently associated with the assessment and do not work effectively when the assessed credit risk criteria interact with each other. To address these issues, we propose MCGDM model for bank credit risk assessment that has two advantages: (1) The imprecise assessment criteria are represented by interval-valued intuitionistic fuzzy sets, and (2) the interactions among the criteria are modeled using fuzzy cognitive maps. When combined with traditional TOPSIS approach to ranking alternatives, we show that the proposed model can be effectively applied to assess bank credit risk.