Machine Learning-Based Fraud Detection in E-Commerce Transactions

Abstract

The rapid growth of e-commerce has heightened fraud risks, demanding advanced fraud detection solutions. This study evaluates five machine learning models Logistic Regression, SVM, KNN, Random Forest, and Gradient Boosting for detecting fraudulent transactions in e-commerce environments. The models were assessed based on accuracy, precision, recall, F1-score, ROC-AUC, and error-related indicators. Results indicate that ensemble-based models, particularly Gradient Boosting and Random Forest, consistently outperform linear models like Logistic Regression, achieving superior balance between precision and recall. Gradient Boosting emerged as the top performer, with the highest accuracy (0.9763), F1-score (0.9765), and ROC-AUC (0.9880), while maintaining a low false negative rate (4.38%). These findings suggest that machine learning models, particularly ensemble methods, provide robust and efficient fraud detection frameworks. The study emphasizes the importance of using recall and F1-score as primary metrics to balance fraud detection sensitivity and operational efficiency.