Modeling the Relationship between Financial Stability and Banking Risks: Artificial Intelligence Approach

This aims to evaluate the effect of financial stability efficiency on the financial stability of banks in Iran and Iraq. In addition, this study aims to identify the main factors affecting financial stability and provide solutions to improve risk management and banking supervision. The statistical population of this research includes 66 banks (22 Iranian banks and 44 Iraqi banks) listed on the stock exchanges of Iran and Iraq. The study period is from 2000 to 2023. A wide range of artificial neural network approaches and machine learning algorithms have been used for data analysis. These methods include artificial neural network, deep neural network, convolutional neural network, recurrent neural network, self-organizing neural network, gradient boosting, random forest, decision tree, spatial clustering, k-means algorithm, k-nearest neighbor, support vector regression and support vector machine. This diversity in analytical methods provides the possibility of comprehensive comparison and evaluation of factors affecting the financial stability of banks. The results of various analyzes show that machine learning models and neural networks have a significant performance in examining and evaluating the financial stability of banks. The artificial neural network showed the highest accuracy with a coefficient of determination of 0.95. Gradient boosting and random forest also had high performance with determination coefficients of 0.9566 and 0.9441. Spatial clustering and k-means algorithms could group banks based on their financial stability with an accuracy of nearly 100%. The variables of capital adequacy ratio, cash flow, bank size, and Z score were identified as the most important factors affecting financial stability. Deep, convolutional, and recurrent neural network models also showed similar performance with coefficients of determination of about 0.94. Support vector regression and support vector machine also provided acceptable results with determination coefficients of 0.9162 and 0.8500. This study emphasizes using artificial intelligence approaches in risk management and banking supervision. The research findings help develop early warning systems, improve banking supervision, and formulate more efficient monetary policies in Iran and Iraq. It is suggested that the monetary authorities of the two countries use these results to revise the capital adequacy rules and strengthen the banks' liquidity management.