Machine Learning-Based Cyber Fraud Detection: A Comparative Study of Resampling Methods for Imbalanced Credit Card Data

The prevalence of online transactions and extensive adoption of credit card payments have contributed to the escalation of credit card cyber fraud in modern society. These trends are propelled by technological advancements, which provide fraudulent actors with more opportunities. Fraudsters exploit victims’ financial vulnerabilities by obtaining illegal access to sensitive credit card information through deceptive means, such as phishing, fraudulent phone calls, and fraudulent SMS messages. This study predicts and detects potential instances of cyber fraud in credit card transactions by employing Machine Learning (ML) techniques, including Decision Tree (DT); Random Forest (RF); Logistic Regression (LR); Support Vector Machine (SVM); K-Nearest Neighbors (KNN); XGBoost; CatBoost; and sampling techniques such as Tomek Link, Synthetic Minority oversampling technique (SMOTE), Edited Nearest Neighbor (ENN), Tomek+ENN, and SMOTE+ENN. To determine the performance of the algorithms in terms of accuracy, precision, recall, F1 score, and ROC-AUC for credit card cyber fraud detection, we conducted a comparative analysis of the extant ML techniques.