Online based Payment Fraud Detection using SMOTE Analysis

Computer programs recognize the importance of studying online payment fraud detection because digital transactions are expanding at high speed. Machine learning provides modern organizations with effective technology to combat fraudulent transactions in digital payment systems. The paper develops an analysis of multiple fraud detection models while focusing on feature selection strategies and model evaluation criteria and optimized fraud detection protocols. The proposed system leverages advanced AI algorithms such as SVM, XGBoost, and Extra Trees Classifier, enhanced through the use of SMOTE (Synthetic Minority Over-sampling Technique) analysis. This integration not only improves detection precision but also significantly reduces false alarm rates. The application of SMOTE has resulted in a 3%–4% increase in accuracy across all algorithms, thereby enhancing their effectiveness and reliability for fraud detection tasks.