mohammadreza etebari

Banks' inability to credit assessment and financial evaluation of customers and forecasting accurately the credit risk of borrowers has devastating effects on the global financial system and economic activity and have been the main causes of global financial crises in recent years.The purpose of this paper is to compile a credit forecasting model for legal customers of private banks by using meta-heuristic algorithms in the branches of Pasargad Bank in the northwest of Iran.

This research is base on the purpose of developmental research and based on the method of performing descriptive work. The statistical population of this study is in two sections of banking experts and legal customers of Pasargad Bank in the northwest of the Iran. The statistical sample size for the first community of 58 banking experts including managers, credit officials and heads of branches in with credit work experience in private banks and for the second community, 427 legal clients were selected based on targeted sampling. In order to collect data in this research, a questionnaire and documents of Pasargad Bank have been used. The validity of the questionnaire was investigated as content validity and based on the indicators of content validity ratio and content validity index. The reliability of the questionnaire was assessed using Cronbach's alpha coefficient. In order to analyze the research data, t-test, confirmatory factor analysis, multilayer neural network, genetically trained neural network, trained neural network with particle swarm optimization and trained neural network with differential evolution will be used.

 The research findings show that all four models are able to predict the credit predictions of the legal customers of private banks and the best way to predict the credit predictions of the legal customers of private banks is the neural network trained with differential evolution algorithm with the least amount of error compared to the other three methods.Originality/Value: In this research by using meta-heuristic algorithms, a new credit forecasting model produce for legal customers of private banks with the least amount of error.