RISK ASSESSMENT IN BANKING TRANSACTIONS USING A FUZZY SYSTEM BASED ON LINGUISTIC RULES AND BEHAVIORAL PARAMETERS

This article presents the development and validation of a fuzzy system for risk assessment in banking transactions, focusing on the detection of suspicious behaviors and potential financial fraud. In light of the increasing digitalization of the banking sector and the growing sophistication of fraudulent practices, a Mamdani-type fuzzy logic approach is proposed, capable of handling uncertainty, imprecise information, and complex behavioral patterns, thereby approximating human reasoning. The system was modeled using eight input variables related to the financial, temporal, and behavioral characteristics of transactions, and one output variable representing the risk level. The linguistic rules were formulated based on expert knowledge and empirical data from reports issued by the Central Bank of Brazil and the Brazilian Federation of Banks. The implementation was carried out in a Python environment using the scikit-fuzzy library. For validation purposes, 4,409 distinct transaction scenarios were simulated, and the results were compared with classifications assigned by a banking risk specialist. The results indicated a high level of agreement between the fuzzy system and human judgment, reaching 82% agreement for five risk classes and 88% when the categories were grouped into three classes. These findings demonstrate that the proposed model is interpretable, robust, and effective as a decision-support tool, contributing to the strengthening of security mechanisms and fraud mitigation in the digital financial environment.