The Impact of Artificial Intelligence-Based Fraud Detection Systems on Transaction Security in Nigerian Commercial Banks

Financial fraud through digital banking channels has escalated significantly in Nigeria, creating substantial losses for commercial banks and eroding public confidence in electronic payment systems. This study empirically investigates the impact of artificial intelligence-based fraud detection systems on transaction security outcomes in Nigerian commercial banks, focusing on institutions operating in Lagos, South West Nigeria. A mixed-methods research design was employed, combining a quantitative analysis of fraud incident data from five commercial banks over a 36-month period with qualitative interviews involving 14 heads of fraud management and IT security. Machine learning model types deployed by each bank were categorised, and their performance metrics including detection accuracy, false positive rates, and incident response times were compared before and after AI system deployment. Findings indicate a mean reduction of 41 percent in successful fraud incidents following AI system implementation, with neural network-based models outperforming rule-based systems in detecting novel attack patterns. However, elevated false positive rates in two banks led to legitimate transaction reversals that generated customer complaints. The study contributes to the emerging literature on AI in financial security by providing longitudinal evidence from a high-fraud-incidence African market. Recommendations include dynamic model retraining protocols and human oversight mechanisms for AI-flagged transaction reviews. Keywords: artificial intelligence, fraud detection, commercial banks, transaction security, Lagos.