Designing an Expert System for Credit Rating of Real Customers of Banks Using Fuzzy Neural Networks

Currently, in Iran’s banking system, nonrepayment of facilities has become one of the biggest issues, and due to the lack of a proper system for proper allocation of facilities, they face a number of problems, including the problem of allocation of loans, the problem of failure to repay loans Of the central bank, or the amount of facilities increased from the amount of reimbursement. The solution of this problem is the credit rating of the customers, which is based on a model based on the theory of fuzzy sets for validation of real customers of the Maskan bank of the East Azerbaijan in Iran in 2016. In this research a structured model was obtained for determination and categorization of input variables for application in the system by factorial analysis then a expert fuzzy system was modelled that consist of six steps. In the first step a fuzzy system is designed that its inputs are financial capacity, support, reliability, repayment record and its outputs is customer credit. In the second step input and outputs are partitioned, in the third step thee partitioned inputs and outputs are converted into fuzzy numbers. The fuzzy inference is compiled in step four. In step five the defuzzifier is conducted. Finally the designed model is tested in step six. These results indicate research model efficiency compared to bank credit measuring experts that they predicate applicants performance according their judgment and intuition.