Credit Card Fraud Detection System Using Gradient Boosting with Deep Feature Engineering and Ensemble Stacking

Abstract

Credit card fraud detection remains a critical challenge due to highly imbalanced transaction data and evolving fraudulent behaviors. This paper presents an effective Credit Card Fraud Detection System using gradient boosting models integrated with deep feature engineering and ensemble stacking. The proposed framework begins with data acquisition and preprocessing to handle missing values and class imbalance, followed by advanced feature engineering to extract meaningful transactional patterns. Multiple gradient boosting classifiers, namely XGBoost, LightGBM, and CatBoost, are trained independently to capture complex non-linear relationships in the data. Their predictions are combined using an ensemble stacking strategy to enhance overall detection performance. The proposed system is evaluated on a real-world credit card transaction dataset containing 284,807 transactions, with a fraud rate of 0.173%. Experimental results demonstrate that the model achieves an accuracy of 99.94%, precision of 0.8472, recall of 0.8243, F1-score of 0.8356, and a ROC–AUC score of 0.9657. The confusion matrix further confirms effective fraud detection with 61 true positives, 11 false positives, and 13 false negatives. These results highlight the robustness of the proposed ensemble-based approach in accurately detecting fraudulent transactions while minimizing false alarms, making it suitable for real-time financial fraud detection systems.