Year: 2026
Type: article
Abstract: Credit card fraud is a major challenge for financial institutions, merchants, and customers because fraudulent transactions can result in significant financial losses and reduced customer confidence. The rapid growth of digital payments has further increased the volume and complexity of financial transactions, making automated fraud detection an important research problem. Machine learning techniques can identify suspicious transaction patterns and support real-time fraud prevention; however, the highly imbalanced nature of fraud datasets makes conventional classification accuracy insufficient for evaluating model performance. This paper presents an automated machine learning framework for credit card fraud detection that addresses class imbalance through sampling strategies and evaluates multiple classification approaches. The original implementation uses data preprocessing, exploratory analysis, statistical analysis, random under-sampling, train-test splitting, and Logistic Regression. The original experimental implementation achieved an accuracy of 93.91%. To improve the methodology and address the limitations of a single-model accuracy-based evaluation, the revised framework incorporates Logistic Regression, Random Forest, XGBoost, and Autoencoder-based anomaly detection. Random under-sampling and Synthetic Minority Over-sampling Technique (SMOTE) are considered for handling class imbalance. Model performance is evaluated using Precision, Recall, F1-score, ROC-AUC, and Precision-Recall AUC (PR-AUC), rather than relying only on overall accuracy. The proposed comparative framework provides a systematic approach for identifying an appropriate fraud detection model while considering the trade-off between fraudulent transactions detected and false-positive alerts generated. The framework can be extended to real-time transaction monitoring and adaptive fraud detection systems.
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204249297