Document Type : Article-Based Dissertations

Authors

• Mehrdad Jeyhoonipour ¹
• somayeh azami ²
• sohrab Delangizan ¹

¹ Razi University

² razi university

Abstract

One of the most significant consequences of financial intermediation activities in banks and credit institutions is exposure to credit risk. As the society is growing and developing, the amount of facilities and liquidity circulation in it increases and the importance of credit health becomes more necessary. Therefore, evaluating and managing credit risk is a vital thing for banks and is also an important solution for implementing banking policies and business strategies. In this research, we intend to design a model based on Bayesian networks in order to predict the credit risk of credit applicants of banks and credit institutions and to increase the efficiency of the services of banks and credit institutions through the automation of decision making to granting credit. In other words, this research aims to provide an effective decision-making system for banks and credit institutions, identify non-committed borrowers and reduce their proportion through the use of Bayesian Networks Model. After determining the most important factors affecting defaults in the form of demographic, socio-economic, financial and credit indicators, by referring to experts and using Dimatel's decision-making technique as one of the methods of identifying causal relationships between factors The main credit risk in the banking system, the conditional relationships of dependence between the variables that explain the default have been set (structural learning), and then parametric learning has been used to detect the conditional probabilities of customer default. The parameters are estimated based on the data of microcredit of natural persons obtained from a Credit Institution in Iran. Bayesian network analysis shows that monthly repayment amount, type of credit, borrower's loan balance in the banking system, age, gender, annual income, and account average are the important variables of default for Explanation the probability of default is more than 70%, respectively. Using the results of this research, the output of the model can be implemented in bank branches in a way that by submitting a request for a facility by a new applicant and taking into account the amount and duration of his requested credit, the probability of default, credit rating and determined the credit that can be allocated to him using the bank's credit policies. Regarding the facilities paid by the bank, it is possible to measure the credit risk of the bank's portfolio using the Bayesian network model and use it for the future credit policy.

Keywords

• Bayesian network
• Credit scoring
• Parametric Learning
• Payment default
• Structural Learning

Main Subjects

• Monetary economics and banking