Improving Accuracy and Efficiency of Online Payment Fraud Detection and Prevention with Machine Learning Models

These days, many incidents of internet fraud are handled by cyber forensics. The likelihood of online fraud is being amplified by the widespread use of the internet. Automated fraud detection in online transactions is a challenging task, as fraudsters are constantly developing new and sophisticated methods. This study focuses on improving an accuracy and efficiency of online payment fraud detection and prevention by integrating advanced data preprocessing, feature extraction, and model optimisation techniques. A robust dataset preprocessing pipeline, including handling missing values, outlier removal, data standardisation, and balancing through undersampling, ensures high-quality input. Key features are extracted to enhance model interpretability and efficiency. Several ML models, including LR, SVM, KNN, and CNN, are employed to classify transactions. Models are assessed by calculating their F1-score, accuracy, precision, and recall. By obtaining an astounding 95% accuracy, 97.72% precision, 99.41% recall, and 98.56% F1 score, the CNN model surpasses conventional ML approaches to provide better outcomes. These results highlight CNN's superiority in capturing complex fraud patterns and maintaining high performance across all metrics. The proposed approach offers a robust solution for real-time fraud detection in online payment systems, ensuring accuracy, efficiency, and scalability.