KMAN Publication Inc. AI and Tech in Behavioral and Social Sciences In Press 2026 10 01 Modeling the Factors Influencing Bank Failure Using a System Dynamics Approach and Genetic Algorithm Modeling the Factors Influencing Bank Failure Using a System Dynamics Approach and Genetic Algorithm EN Journal Article <p>The present study is applied in terms of its purpose and employed a mixed qualitative and quantitative approach. The factors affecting bank failure were identified in the qualitative phase through a comprehensive review of the theoretical literature and semi-structured interviews with experts. The participants consisted of 11 senior bank managers, researchers, and university professors who were selected using non-probability purposive and convenience sampling. The data included information on capital, profit, accumulated losses, and other bank-related variables for Bank Melli Iran during 2011–2024. First, a system dynamics model was developed to represent the relationships among resource mobilization, funding composition, credit efficiency, lending income, funding costs, profit or loss, capital, and accumulated losses, and was evaluated using historical data. Subsequently, a genetic algorithm was employed to identify the minimum level of intervention, and two policy scenarios were examined. In the first scenario, three policy levers—resource mobilization, funding composition, and credit efficiency—were optimized simultaneously. In the second scenario, the final effectiveness of the policy levers was constrained to a maximum level of one, reflecting the assumption that intervention should not rely on the market share of other banks. The results showed that, within the structure of the model and the defined policy range, the multi-lever intervention scenario provided a feasible solution for satisfying the constraint over 13 transition steps, with the minimum accumulated-loss-to-capital ratio reaching approximately −0.49996. In contrast, under the scenario in which the final effectiveness was constrained to a level of one, no feasible solution was observed during 2016–2019, and even the maximum permissible intervention was insufficient to satisfy the constraint. These findings indicate that, under the assumptions of the model, the capacity for internal banking reform may not, by itself, be sufficient to control the capital-crisis indicator during certain critical periods. Sensitivity analysis showed that the model output was more sensitive to inflation, while credit efficiency generated the greatest sensitivity among the policy levers in the trajectory of the accumulated-loss-to-capital ratio. Compared with historical data, the baseline model achieved a correlation coefficient of approximately 0.92 and a mean error of less than 9%. Overall, the integration of system dynamics and genetic algorithms, within the data, model structure, policy-lever range, and assumptions of the study, provides a framework for analyzing the temporal trajectory of capital crises and comparing intervention scenarios in Bank Melli Iran. However, the findings should not be interpreted as a definitive guarantee of preventing bank failure; rather, they represent an estimate of the relative capacity of alternative intervention pathways to control an operational indicator of capital distress.</p>