A Comprehensive Comparative Study of Machine Learning Algorithms for Fraud Detection in Financial Transactions

Abstract The rapid expansion of digital financial services has significantly increased the risk and sophistication of fraudulent activities. Traditional rule-based fraud detection systems lack adaptability against evolving fraud patterns and often fail in highly imbalanced data environments. Machine learning (ML) techniques provide automated, scalable, and adaptive solutions for fraud detection. This study presents a comprehensive comparative analysis of five prominent ML algorithms—Logistic Regression, Decision Tree, Support Vector Machine (SVM), Random Forest, and Extreme Gradient Boosting (XGBoost)—for financial fraud detection. The models are evaluated using multiple performance metrics, including accuracy, precision, recall, F1-score, and AUC-ROC, under standardized preprocessing and class imbalance handling strategies. Experimental findings indicate that ensemble and boosting methods outperform traditional classifiers, particularly in recall and AUC metrics, making them more suitable for real-world financial systems. The study further analyzes interpretability, computational complexity, and deployment feasibility to provide practical recommendations. Keywords : Fraud Detection, Machine Learning, Random Forest, XGBoost, Support Vector Machine, Class Imbalance, Financial Security, AUC-ROC.