Explainable AI for Enhanced Anomaly Detection in Fraud Detection
محورهای موضوعی : Machine learningReza Amiri 1 , Mohammad Hadi Zahedi 2 , Mehdi Azadimotlagh 3
1 - جهاد دانشگاهی
2 - دانشگاه صنعتی خواجه نصیرالدین طوسی
3 - دانشگاه خلیج فارس
کلید واژه: Explainable Artificial Intelligence, Anomaly Detection, Fraud Detection, Interpretable Models, Machine Learning.,
چکیده مقاله :
Abstract: The application of machine learning has become indispensable in the critical domain of financial fraud detection. However, a major limitation of traditional models is their "black box" nature, which obscures the reasoning behind a flagged transaction. This lack of transparency often leads to many false positives, which can undermine customer trust and incur substantial operational expenses. To address this challenge, this paper proposes a novel framework for Explainable Anomaly Detection in financial fraud, using advanced Explainable AI (XAI) techniques to provide clear insights into the model's predictive processes. Our approach is designed to move beyond a simplistic binary output of "fraud/no fraud." Our framework combines advanced anomaly detection models, like Isolation Forests and Deep Auto-encoders with model-agnostic explanation methods such as SHAP and LIME, to clearly show which features contribute to a transaction’s anomaly score. The efficacy of our framework has been evaluated using a financial transaction benchmark dataset. The results show that integrating XAI not only makes the system more transparent and trustworthy, but also improves the efficiency of fraud investigations. Based on these results, our method reduces the time and resources needed for manual reviews, while still maintaining high accuracy in detecting fraudulent activities.
Abstract: The application of machine learning has become indispensable in the critical domain of financial fraud detection. However, a major limitation of traditional models is their "black box" nature, which obscures the reasoning behind a flagged transaction. This lack of transparency often leads to many false positives, which can undermine customer trust and incur substantial operational expenses. To address this challenge, this paper proposes a novel framework for Explainable Anomaly Detection in financial fraud, using advanced Explainable AI (XAI) techniques to provide clear insights into the model's predictive processes. Our approach is designed to move beyond a simplistic binary output of "fraud/no fraud." Our framework combines advanced anomaly detection models, like Isolation Forests and Deep Auto-encoders with model-agnostic explanation methods such as SHAP and LIME, to clearly show which features contribute to a transaction’s anomaly score. The efficacy of our framework has been evaluated using a financial transaction benchmark dataset. The results show that integrating XAI not only makes the system more transparent and trustworthy, but also improves the efficiency of fraud investigations. Based on these results, our method reduces the time and resources needed for manual reviews, while still maintaining high accuracy in detecting fraudulent activities.
References
[1] Ali, A., Abd Razak, S., Othman, S. H., Eisa, T. A. E., Al-Dhaqm, A., Nasser, M., ... & Saif, A. (2022). Financial fraud detection based on machine learning: a systematic literature review. Applied Sciences, 12(19), 9637.
[2] Zhou, Y., Li, H., Xiao, Z., & Qiu, J. (2023). A user-centered explainable artificial intelligence approach for financial fraud detection. Finance Research Letters, 58, 104309.
[3] Zhang, Y., Xu, T., Song, X., Zhu, X. F., Feng, Z., & Wu, X. J. (2024). Towards accurate unsupervised video captioning with implicit visual feature injection and explicit. Pattern Recognition Letters, 183, 133-139.
[4] Li, Z., Zhu, Y., & Van Leeuwen, M. (2023). A survey on explainable anomaly detection. ACM Transactions on Knowledge Discovery from Data, 18(1), 1-54.
[5]. Iqbal, A., & Amin, R. (2025). An efficient mechanism for time series forecasting and anomaly detection using explainable artificial intelligence. The Journal of Supercomputing, 81(4), 523.
[6] Dhanesha, P., & Mehta, D. (2024, December). Uncovering Hidden Frauds: Isolation Forest-Based Anomaly Detection in Credit Card Transactions. In International Conference on Information and Communication Technology for Competitive Strategies (pp. 39-51). Singapore: Springer Nature Singapore.
[7] Neloy, A. A., & Turgeon, M. (2024). A comprehensive study of auto-encoders for anomaly detection: Efficiency and trade-offs. Machine Learning with Applications, 17, 100572.
[8] Lundberg, S. M., & Lee, S. I. (2017). A unified approach to interpreting model predictions. Advances in neural information processing systems, 30.
[9] Ribeiro, M. T., Singh, S., & Guestrin, C. (2016, August). " Why should i trust you?" Explaining the predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining (pp. 1135-1144).
[10] Dal Pozzolo, A., Boracchi, G., Caelen, O., Alippi, C., & Bontempi, G. (2017). Credit card fraud detection: a realistic modeling and a novel learning strategy. IEEE transactions on neural networks and learning systems, 29(8), 3784-3797.
[11] Agrawal, V., Panigrahi, B. K., & Subbarao, P. M. V. (2018). Increasing reliability of fault detection systems for industrial applications. IEEE Intelligent Systems, 33(3), 28-39.
[12] Dal Pozzolo, A., Caelen, O., Le Borgne, Y. A., Waterschoot, S., & Bontempi, G. (2014). Learned lessons in credit card fraud detection from a practitioner perspective. Expert systems with applications, 41(10), 4915-4928.
[13] Branco, P., Torgo, L., & Ribeiro, R. P. (2017, April). Relevance-based evaluation metrics for multi-class imbalanced domains. In Pacific-Asia Conference on Knowledge Discovery and Data Mining (pp. 698-710). Cham: Springer International Publishing.