Credit Card Fraud Detection with Cost-Sensitive Machine Learning: A Comparative Application on Open-Access and Severely Imbalanced Transaction Data

Öz

This study examines credit card fraud detection on an open-access transaction dataset within a cost-sensitive machine learning framework. After removing 1,081 duplicate records from the 284,807 transactions in the benchmark dataset shared by the Worldline-ULB group, the analyses were conducted on 283,726 observations and 473 fraudulent transactions. Logistic Regression, Random Forest, Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM) models were compared under a common training-validation-test split; in addition to the V1–V28, Time, and Amount variables, log_amount and hour_bucket were derived. Performance was evaluated using PR-AUC, ROC-AUC, recall, F1 score, and balanced accuracy, while the decision threshold was selected according to the value that minimizes the expected total cost on the validation set. The study presents an extended comparison covering, alongside random splitting, time-based (chronological) splitting, the SMOTE, undersampling, SMOTE-Tomek, and focal loss approaches for class imbalance, Platt and isotonic probability calibration, and the comparison of different LightGBM configurations. On the validation set, the highest PR-AUC was observed in the XGBoost model with 0.8882, whereas the lowest cost was observed in the Random Forest model with 1,844.88. On the test set, XGBoost produced the strongest discriminative performance with a PR-AUC of 0.8207 and a ROC-AUC of 0.9710, while the lowest total cost was obtained by Random Forest with 3,896.58; LightGBM under Configuration B produced a comparable cost of 3,837.01, and focal loss, together with the XGBoost base model, provided an additional gain with a cost of 3,421.29. The findings show that, in severely imbalanced financial data, model selection alone is not sufficient; the decision threshold, the class-imbalance approach, and the cost assumptions can change the model ranking.

Anahtar Kelimeler

• credit card fraud detection
• imbalanced classification
• cost-sensitive learning
• threshold optimization
• XGBoost

Maliyet Duyarlı Makine Öğrenmesi ile Kredi Kartı Dolandırıcılığı Tespiti: Açık Erişimli ve Ağır Dengesiz İşlem Verisi Üzerinde Karşılaştırmalı Bir Uygulama

Öz

Bu çalışma, açık erişimli kredi kartı işlem verisi üzerinde maliyet duyarlı makine öğrenmesi çerçevesinde dolandırıcılık tespitini incelemektedir. Worldline-ULB grubu tarafından paylaşılan karşılaştırma veri kümesindeki 284.807 işlemden 1.081 yinelenen kayıt çıkarılmış, analizler 283.726 gözlem ve 473 dolandırıcılık işlemi üzerinde yapılmıştır. Lojistik Regresyon (Logistic Regression), Rastgele Orman (Random Forest), Aşırı Gradyan Artırma (Extreme Gradient Boosting, XGBoost) ve Hafif Gradyan Artırma Makinesi (Light Gradient Boosting Machine, LightGBM) modelleri ortak eğitim-doğrulama-test bölünmesi altında karşılaştırılmış; V1–V28, Time ve Amount değişkenlerine ek olarak log_amount ve hour_bucket türetilmiştir. Başarım, PR-AUC, ROC-AUC, duyarlılık, F1 skoru ve dengelenmiş doğruluk ile değerlendirilmiş; karar eşiği doğrulama kümesinde beklenen toplam maliyeti en aza indiren değere göre seçilmiştir. Çalışma; rastgele bölme yanında zaman-temelli (kronolojik) bölme, sınıf dengesizliğine yönelik SMOTE, alt örnekleme, SMOTE-Tomek ve focal loss yaklaşımları, Platt ile isotonic olasılık kalibrasyonu ve farklı LightGBM yapılandırmalarının karşılaştırılması üzerinden genişletilmiş bir karşılaştırma sunmaktadır. Doğrulama kümesinde en yüksek PR-AUC 0,8882 ile XGBoost modelinde, en düşük maliyet ise 1.844,88 ile rastgele orman modelinde gözlenmiştir. Test kümesinde XGBoost 0,8207 PR-AUC ve 0,9710 ROC-AUC ile en güçlü ayrıştırma başarımını üretirken, en düşük toplam maliyet 3.896,58 ile rastgele orman tarafından elde edilmiştir; LightGBM Yapılandırma B altında 3.837,01 ile yakın bir maliyet üretmiş, focal loss ise XGBoost taban modeli ile birlikte 3.421,29 maliyetle ek bir kazanım sağlamıştır. Bulgular, ağır dengesiz finansal verilerde model seçiminin tek başına yeterli olmadığını; karar eşiği, sınıf dengesizliği yaklaşımı ve maliyet varsayımlarının model sıralamasını değiştirebildiğini göstermektedir.

Anahtar Kelimeler

• kredi kartı dolandırıcılığı
• dengesiz sınıf yapısı
• maliyet duyarlı öğrenme
• karar eşiği
• XGBoost

Kaynakça

1. Abdul Salam, M., Fouad, K. M., Elbably, D. L. ve Elsayed, S. M. 2024. Federated learning model for credit card fraud detection with data balancing techniques. Neural Computing and Applications, 36(11), 6231–6256.
2. Afriyie, J. K., Tawiah, K., Pels, W. A., Addai-Henne, S., Dwamena, H. A., Owiredu, E. O., Ayeh, S. A. ve Eshun, J. 2023. A supervised machine learning algorithm for detecting and predicting fraud in credit card transactions. Decision Analytics Journal, 6, 100163.
3. Ahmed, K. H., Axelsson, S., Li, Y. ve Sagheer, A. M. 2025. A credit card fraud detection approach based on ensemble machine learning classifier with hybrid data sampling. Machine Learning with Applications, 20, 100675.
4. Alarfaj, F. K., Malik, I., Khan, H. U., Almusallam, N., Ramzan, M. ve Ahmed, M. 2022. Credit card fraud detection using state-of-the-art machine learning and deep learning algorithms. IEEE Access, 10, 39700–39715.
5. Albalawi, T. ve Dardouri, S. 2025. Enhancing credit card fraud detection using traditional and deep learning models with class imbalance mitigation. Frontiers in Artificial Intelligence, 8, 1643292.
6. Altan, G. ve Zafer, M. R. 2024. Denetimli makine öğrenmesi yöntemleri ile kredi kartı sahteciliğini tahmin etme: karşılaştırmalı analiz. İktisat Politikası Araştırmaları Dergisi, 11(2), 242–262.
7. Bahnsen, A. C., Aouada, D., Stojanovic, A. ve Ottersten, B. 2016. Feature engineering strategies for credit card fraud detection. Expert Systems with Applications, 51, 134–142.
8. Baisholan, N., Dietz, J. E., Gnatyuk, S., Turdalyuly, M., Matson, E. T. ve Baisholanova, K. 2025. FraudX AI: An interpretable machine learning framework for credit card fraud detection on imbalanced datasets. Computers, 14(4), 120.

9. Batista, G. E. A. P. A., Prati, R. C. ve Monard, M. C. 2004. A study of the behavior of several methods for balancing machine learning training data. ACM SIGKDD Explorations Newsletter, 6(1), 20–29.
10. Breiman, L. 2001. Random forests. Machine Learning, 45(1), 5–32.
11. Btoush, E., Zhou, X., Gururajan, R., Chan, K. C. ve Alsodi, O. 2025. Achieving excellence in cyber fraud detection: A hybrid ML+DL ensemble approach for credit cards. Applied Sciences, 15(3), 1081.
12. C-Rella, J., Cao, R. ve Vilar, J. M. 2024. Cost-sensitive thresholding over a two-dimensional decision region for fraud detection. Information Sciences, 657, 119956.
13. Chawla, N. V., Bowyer, K. W., Hall, L. O. ve Kegelmeyer, W. P. 2002. SMOTE: Synthetic minority over-sampling technique. Journal of Artificial Intelligence Research, 16, 321–357.
14. Chen, T. ve Guestrin, C. 2016. XGBoost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794.
15. Cherif, A., Badhib, A., Ammar, H., Alshehri, S., Kalkatawi, M. ve Imine, A. 2023. Credit card fraud detection in the era of disruptive technologies: A systematic review. Journal of King Saud University – Computer and Information Sciences, 35(1), 145–174.
16. Chhabra, R., Goswami, S. ve Ranjan, R. K. 2024. A voting ensemble machine learning based credit card fraud detection using highly imbalance data. Multimedia Tools and Applications, 83(18), 54729–54753.
17. Dal Pozzolo, A., Boracchi, G., Caelen, O., Alippi, C. ve Bontempi, G. 2018. Credit card fraud detection: A realistic modeling and a novel learning strategy. IEEE Transactions on Neural Networks and Learning Systems, 29(8), 3784–3797.
18. Fanai, H. ve Abbasimehr, H. 2023. A novel combined approach based on deep Autoencoder and deep classifiers for credit card fraud detection. Expert Systems with Applications, 217, 119562.
19. Friedman, J. H. 2001. Greedy function approximation: A gradient boosting machine. The Annals of Statistics, 29(5), 1189–1232.
20. Ghosh Dastidar, K., Jurgovsky, J., Siblini, W. ve Granitzer, M. 2022. NAG: Neural feature aggregation framework for credit card fraud detection. Knowledge and Information Systems, 64(3), 831–858.
21. Güney, İ. ve Selvi, İ. H. 2023. Makine Öğrenmesi Yöntemleriyle Anormal İçme Suyu Tüketimlerinin Tespit Edilmesi ve Tahmin Modellerinin Geliştirilmesi. Journal of Intelligent Systems: Theory and Applications, 6(2), 159-173.
22. Ileberi, E., Sun, Y. ve Wang, Z. 2022. A machine learning based credit card fraud detection using the GA algorithm for feature selection. Journal of Big Data, 9, 24.
23. Jurgovsky, J., Granitzer, M., Ziegler, K., Calabretto, S., Portier, P.-E., He-Guelton, L. ve Caelen, O. 2018. Sequence classification for credit-card fraud detection. Expert Systems with Applications, 100, 234–245.
24. Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q. ve Liu, T.-Y. 2017. LightGBM: A highly efficient gradient boosting decision tree. Advances in Neural Information Processing Systems, 30, 3146–3154.
25. Lin, T.-Y., Goyal, P., Girshick, R., He, K. ve Dollár, P. 2017. Focal loss for dense object detection. Proceedings of the IEEE International Conference on Computer Vision, 2999–3007.
26. Motie, S. ve Raahemi, B. 2024. Financial fraud detection using graph neural networks: A systematic review. Expert Systems with Applications, 240, 122156.
27. Niculescu-Mizil, A. ve Caruana, R. 2005. Predicting good probabilities with supervised learning. Proceedings of the 22nd International Conference on Machine Learning, 625–632.
28. Platt, J. C. 1999. Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods. In A. J. Smola, P. Bartlett, B. Schölkopf ve D. Schuurmans (Eds.), Advances in Large Margin Classifiers, MIT Press, 61–74.
29. Randhawa, K., Loo, C. K., Seera, M., Lim, C. P. ve Nandi, A. K. 2018. Credit card fraud detection using AdaBoost and majority voting. IEEE Access, 6, 14277–14284.
30. Sinap, V. 2024. Comparative analysis of machine learning techniques for credit card fraud detection: Dealing with imbalanced datasets. Turkish Journal of Engineering, 8(2), 196–208.
31. Tang, Y. ve Liang, Y. 2024. Credit card fraud detection based on federated graph learning. Expert Systems with Applications, 256, 124979.
32. Tayebi, M. ve El Kafhali, S. 2025. A novel approach based on XGBoost classifier and Bayesian optimization for credit card fraud detection. Cyber Security and Applications, 3, 100093.
33. Tekin, R. ve Yaman, O. 2023. Akıllı Ev Sistemleri için XGBoost Tabanlı Saldırı Tespit Yöntemi. Journal of Intelligent Systems: Theory and Applications, 6(2), 152-158.
34. Xie, Y., Liu, G., Yan, C., Jiang, C., Zhou, M. C. ve Li, M. 2024. Learning transactional behavioral representations for credit card fraud detection. IEEE Transactions on Neural Networks and Learning Systems, 35(4), 5735–5748.
35. Yıltırak, İ. ve Öztürk, A. 2025. Nesnelerin İnterneti Saldırılarının Tespitinde Makine Öğrenmesi Algoritmalarının Performanslarının Karşılaştırılması. Journal of Intelligent Systems: Theory and Applications, 8(2), 130-140.
36. Zadrozny, B. ve Elkan, C. 2002. Transforming classifier scores into accurate multiclass probability estimates. Proceedings of the Eighth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 694–699.