Future of Work and Digital Management Journal 3 Serial Number 10 2025 10 05 Identification of Agents in an Agent-Based Simulation Model for Fraud Detection in the Banking System to Develop Scenarios Identification of Agents in an Agent-Based Simulation Model for Fraud Detection in the Banking System to Develop Scenarios 1 12 EN Journal Article 2025 07 07 <table> <tbody> <tr> <td> <p>The objective of this study is to design and validate an agent-based simulation model for fraud detection in the banking credit ecosystem by identifying core agents, defining their attributes and behaviors, and testing system resilience under multiple economic scenarios. This research employed a qualitative–computational approach. Expert interviews with banking professionals and academic specialists were conducted to identify key agents, attributes, and relationships within the credit system. The agent-based simulation was then designed around three core agents: Customers, Banks, and the Central Bank. Each agent was assigned distinct financial and behavioral attributes, and their interactions were formalized through computational rules. Customer repayment behavior was modeled by combining financial capacity indicators (income, debt-to-income ratio, FICO score, revolving utilization) with behavioral intention (willingness to pay). Banks adapted their lending policies based on observed defaults, while the Central Bank influenced the system through macroeconomic variables such as inflation and unemployment. Finally, scenario analysis was applied to test the model under optimistic, realistic, and pessimistic macroeconomic conditions. The findings indicated that repayment decisions are were primarily influenced by the combination of behavioral intention (weighted at 70%) and financial capacity (weighted at 30%). Inflation and unemployment substantially reduced repayment capacity, confirming the importance of macroeconomic pressures in fraud emergence. Banks demonstrated adaptive responses by tightening or loosening credit policies depending on observed defaults. Scenario analysis revealed that in optimistic environments defaults were minimal; in realistic conditions, defaults were moderate and policy adjustments were balanced; and in pessimistic environments, defaults rose sharply, forcing restrictive credit strategies. The agent-based model successfully captured the dynamic interplay of individual behavior, institutional policy, and macroeconomic shocks in fraud detection within the credit system. Its robustness across scenarios demonstrates the utility of agent-based simulation as a complement to machine-learning methods, offering a tool for banks and regulators to anticipate fraud risks and test policy interventions under varying economic conditions.</p> </td> </tr> </tbody> </table> https://www.journalfwdmj.com/index.php/fwdmj/article/download/102/130