Real-Time Credit Card Fraud Detection Using Ensemble and Supervised Learning Approaches

Background: Credit card fraud has been a growing concern with the expansion of digital payment systems. Traditional fraud detection methods face challenges in adapting to new fraudulent patterns and often result in high rates of false positives (FP) and false negatives FN). Machine learning (ML) offers a promising solution by learning from historical data to detect hidden patterns within transactions.  Objectives: The goal of this project is to create a system for the real-time identification of fraudulent credit card transactions that is powered by machine learning., with a focus on reducing false negatives and false positives.  Methods: Several ML methods were used in this study, for analyzing transaction data, including Random Forest, Voting Classifier, Logistic Regression, Decision Tree, and XGBoost. The dataset used for training and validation was obtained from publicly available credit card transaction data, focusing on recognizing key characteristics that indicate potential fraudulent behavior.  Results: The machine learning model exhibited higher performance over traditional rule-based systems., achieving an accuracy rate of 98%, with a significant reduction in both false positives and false negatives. With respective area under the receiver operating characteristic (ROC) curves of 99.14% and 99.13%, the XGBoost and Voting Matrix models performed the best.  Conclusion: This study shows that ML algorithms can significantly improve the identification of credit card fraud, offering a more flexible and precise system in contrast to conventional approaches.