<![CDATA[The rapid growth of digital transactions has led to an increase in fraudulent activities within online payment systems. Traditional fraud detection methods based on rule-based systems have limitations in identifying evolving fraud patterns. This study aims to analyze the performance of various Machine Learning (ML) models in detecting fraudulent transactions using the Online Payment Fraud Detection Dataset from Kaggle. The models tested include Logistic Regression, Decision Tree, Random Forest, XGBoost, Support Vector Machine (SVM), and Bidirectional Long Short-Term Memory (BiLSTM). The study also evaluates the impact of data balancing techniques, namely Synthetic Minority Oversampling Technique (SMOTE) and Random Undersampling, on model performance. The results indicate that XGBoost and BiLSTM achieved the highest F1-scores of 88% and 90%, respectively, with SMOTE significantly improving recall rates. These findings suggest that ML can be effectively applied to financial security systems, with XGBoost being more suitable for real-time fraud detection, while BiLSTM excels in identifying complex transaction patterns. Future research should focus on optimizing computational efficiency and exploring Explainable AI techniques to enhance model interpretability.]]>
Copyrights © 2025
