Securing Digital Transactions: Machine Learning for Credit Card Fraud Detection
The rapid growth of digital payments has increased the importance of reliable credit card fraud detection systems. Traditional rule-based approaches can struggle to identify com plex and evolving fraud patterns, particularly when fraudulent transactions represent only a small proportion of transaction data. This paper presents a machine learning-based framework that combines data preprocessing, class-imbalance handling, autoencoder-based feature representation, and XGBoost clas sification. The study uses the publicly available credit card transaction dataset containing 284,807 transactions, including 492 fraudulent transactions. Robust scaling is used to reduce the influence of extreme values, while the Synthetic Minority Oversampling Technique (SMOTE) is applied only to the training data. An autoencoder learns a compact representation of trans action features and calculates reconstruction error; these outputs are combined with the prepared features for downstream clas sification. XGBoost is used to distinguish legitimate transactions from potentially fraudulent ones. The framework is evaluated using precision, recall, F1-score, area under the precision recall curve (PR-AUC), and area under the receiver operating characteristic curve (ROC-AUC). The paper also describes an API-oriented deployment workflow. Numerical results should be reported from the final reproducible experiment; illustrative or previously obtained values must not be treated as independently verified without rerunning the implementation.
Authors
- Deepak Hosamane M
- Dhanush M.S
- Vinutha H K
- Darshan B.A
- Jeevan S.S
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-13
- DOI
- https://doi.org/10.5281/zenodo.24200166
- Primary Topic
- Imbalanced Data Classification Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00