Investigating risk management and performance of customer ranking decisions in the banking industry Using a comparison of applied mathematical models(Case study: Bank Saderat)

Some of the benefits of using credit rating models and implementing the accreditation process are to help increase the cash flow in the bank, make better decisions about granting facilities and increase the assurance of repayment of facilities. Therefore, in this study, with the aim of selecting the optimal model and effective variables for credit rating of bank customers. For this purpose, ۶۰۴ real customers of Bank Saderat in the period ۲۰۱۰-۲۰۲۰ were selected. Of these, ۳۰۵ are among the good customers and ۲۹۹ are among the bad customers. In the first stage, ۹ variables were identified as effective primary variables, some of which were identified and eliminated by models as ineffective variables in customers' credit status. The models used in this research are: neural networks with error propagation algorithm, GMDH networks, neural networks with radius axis algorithm, logit model, probit model, audit analysis model). The results of comparison between these models showed that neural networks with radius-based algorithm and GMDH neural networks have the highest accuracy in predicting the credit behavior of bank customers.