Neuro-Fuzzy Inference System Application for Credit Rating of Bank Customers



Nowadays, credit risk is recognized as one of the most important bankruptcy factors of banks and financial institutions. In order to manage and control this risk, design of credit rating models is undeniable necessity. Credit rating is used to identify the probability of credit default and on the other side classify the customers into two groups: good and bad accounts. Until now, various statistical methods, including discriminant analysis, logistic regression and neural networks have been developed for credit rating. Meanwhile, neural networks due to the high flexibility and precision, in recent years have been considered more. This study presents a credit rating model using adaptive neuro-fuzzy inference system to predict financial performance of bank customers. In this model, debt ratio, operational ratio, and return on equity have been selected as input variables, and on the other side the probability of default has been considered as output variable. After training and testing the model based on data from Keshavarzi bank over 2001-2006, the model predicts the credit status of customers with Accuracy of 69.36%.
JEL classification: C2; G240