Book

Description

Credit card fraud has become a significant challenge for financial institutions as digital transactions continue to expand. Detecting fraudulent activities accurately while minimizing false alarms requires intelligent, transparent, and economically efficient decision-making systems. This research monograph presents an integrated framework for credit card fraud detection and prevention using machine learning, explainable artificial intelligence, and cost-sensitive decision analysis. Drawing on a dataset of 284,807 credit card transactions, including 492 fraudulent cases, the study evaluates Logistic Regression, Random Forest, and XGBoost under severe class imbalance. It examines classification performance, alternative imbalance treatments, cost-sensitive decision thresholds, and SHAP-based model explanations. The findings demonstrate that strong predictive performance does not necessarily guarantee operational effectiveness. While XGBoost achieves competitive results under random validation, Random Forest demonstrates greater robustness in the chronological evaluation. The analysis also highlights the importance of balancing fraud detection against false-positive alerts and associated operational costs. By integrating predictive analytics, explainability, risk-based intervention, and human oversight, this monograph offers a decision-support perspective on intelligent fraud prevention. The book is intended for researchers, banking professionals, financial technology specialists, risk managers, and postgraduate students interested in artificial intelligence, cybersecurity, and financial innovation.