DESIGN AND IMPLEMENTATION OF A CREDIT CARD FRAUD DETECTION SYSTEM

The rapid growth of online financial transactions has led to a corresponding rise in credit card fraud, posing serious risks to consumers and financial institutions. Although modern security mechanisms such as encryption, tokenization, and multi-factor authentication have strengthened digital payment platforms, fraudsters still exploit system vulnerabilities especially in card-not-present transactions where physical verification is impossible. This project develops a machine learning-based credit card fraud detection system capable of identifying and preventing fraudulent activities in real time. The system analyzes historical and live transaction data to detect unusual behavioral patterns that deviate from a user’s normal spending profile. Machine learning techniques including anomaly detection, logistic regression, and random forest classification are applied to classify transactions as legitimate or fraudulent. A dataset consisting of 284,807 transactions, including 492 confirmed fraud cases (0.17%), was used to train and evaluate the models. Statistical results show that the final model achieved an accuracy of 99.84%, a precision of 92.15%, and a recall of 86.42%, indicating strong capability in correctly identifying fraudulent activities while minimizing false alarms. The area under the ROC curve (AUC) was 0.98, demonstrating robust discriminative performance. The outcomes of this study include a significant reduction in fraud-related risks through early detection and automated alert mechanisms. The findings also provide insight into the statistical behavior of fraudulent transactions and the effectiveness of machine learning in financial security. Overall, this project contributes to the advancement of secure, intelligent, and reliable online payment systems, offering a foundation for future improvements in financial cybersecurity and fraud analytics.