FEATURE ENGINEERING AND EXPLAINABLE MACHINE LEARNING FOR CREDIT CARD FRAUD DETECTION: EVIDENCE FROM TEMPORAL TRANSACTION DATA
- Authors
-
-
Mahmudova Dilnoza Avaz qizi
“Business Development Bank” IT, 1st Category Specialist
-
- Keywords:
- credit card fraud detection feature engineering transaction history XGBoost explainable machine learning SHAP imbalanced classification
- Abstract
- Credit card fraud detection is challenging because individual transaction attributes may not reveal whether current activity is unusual relative to previous behavior. This study evaluates an expanded feature representation for XGBoost under temporal holdout evaluation. Two identically configured models were compared on 1,852,394 transactions: a baseline using four original variables and an engineered model adding ten historical, temporal, geographic, demographic, and merchant-related features. Average Precision increased from 0.6762 to 0.9299, a difference of +0.2537 (paired class-stratified bootstrap 95% CI: [0.2349, 0.2712]). F1-score, MCC, precision, recall, ROC-AUC, and Recall@5% also improved. SHAP ranked transaction amount, seven-day historical average amount, transaction hour, 24-hour transaction frequency, and city population highest. Permutation analysis supported the predictive utility of the current-amount-to-historical-average ratio. Historical features excluded current and simultaneous transactions; test merchant risk used training labels only. The findings support richer feature representation in the evaluated setting. However, all ten engineered features were introduced jointly, preventing attribution of the total improvement to transaction-history features or any single variable. Single-holdout evaluation and unmodeled fraud-label delays limit generalization.
- References
-
[1] A. Dal Pozzolo, O. Caelen, Y.-A. Le Borgne, S. Waterschoot, and G. Bontempi, ‘Learned Lessons in Credit Card Fraud Detection from a Practitioner Perspective’, 2014.
[2] A. Dal Pozzolo, G. Boracchi, O. Caelen, C. Alippi, and G. Bontempi, ‘Credit Card Fraud Detection: A Realistic Modeling and a Novel Learning Strategy’, 2018.
[3] C. Whitrow, D. J. Hand, P. Juszczak, D. Weston, and N. M. Adams, ‘Transaction Aggregation as a Strategy for Credit Card Fraud Detection’, 2009.
[4] A. Correa Bahnsen, D. Aouada, A. Stojanovic, and B. Ottersten, ‘Feature Engineering Strategies for Credit Card Fraud Detection’, 2016.
[5] V. Van Vlasselaer et al., ‘APATE: A Novel Approach for Automated Credit Card Transaction Fraud Detection Using Network-Based Extensions’, 2015.
[6] J. Jurgovsky et al., ‘Sequence Classification for Credit-Card Fraud Detection’, 2018.
[7] J. Davis and M. Goadrich, ‘The Relationship Between Precision-Recall and ROC Curves’, 2006.
[8] T. Saito and M. Rehmsmeier, ‘The Precision-Recall Plot Is More Informative than the ROC Plot When Evaluating Binary Classifiers on Imbalanced Datasets’, 2015.
[9] D. Chicco and G. Jurman, ‘The Advantages of the Matthews Correlation Coefficient (MCC) over F1 Score and Accuracy in Binary Classification Evaluation’, 2020.
[10] T. Chen and C. Guestrin, ‘XGBoost: A Scalable Tree Boosting System’, 2016.
[11] S. M. Lundberg and S.-I. Lee, ‘A Unified Approach to Interpreting Model Predictions’, 2017.
[12] W. Hilal, S. A. Gadsden, and J. Yawney, ‘Financial Fraud: A Review of Anomaly Detection Techniques and Recent Advances’, 2022.
[13] A. Cherif, A. Badhib, H. Ammar, S. Alshehri, M. Kalkatawi, and A. Imine, ‘Credit Card Fraud Detection in the Era of Disruptive Technologies: A Systematic Review’, 2023.
[14] Y.-Y. Hsin, T.-S. Dai, Y.-W. Ti, M.-C. Huang, T.-H. Chiang, and L.-C. Liu, ‘Feature Engineering and Resampling Strategies for Fund Transfer Fraud with Limited Transaction Data and a Time-Inhomogeneous Modi Operandi’, 2022.
[15] V. Van Vlasselaer et al., ‘APATE: A Novel Approach for Automated Credit Card Transaction Fraud Detection Using Network-Based Extensions’, 2015.
[16] T. Saito and M. Rehmsmeier, ‘The Precision-Recall Plot Is More Informative than the ROC Plot When Evaluating Binary Classifiers on Imbalanced Datasets’, 2015.