Adaptive Feature Regrouping and Community Detection for Robust Visualization in Fraud Detection Systems
Keywords:
Fraud Detection, Explainable Artificial Intelligence (XAI), Adversarial Robustness, Graph-Based Community Detection, Financial Transaction AnalyticsAbstract
Fraud detection systems often rely on static visualizations and rule-based explanations that are vulnerable to adversarial manipulation of input features. We propose a novel framework that replaces conventional decision-path visualization with an adaptive feature regrouping module designed to produce robust, interpretable outputs even under data poisoning attacks. The core methodology operates in two stages: first, an anomaly-driven importance sampling mechanism selects transactions exhibiting statistically significant distributional shifts from historical baselines, thereby focusing on high-impact and potentially adversarial instances. Second, we construct a bipartite graph from user-merchant interactions within this sampled subset and apply modularity-optimized community detection using the Louvain algorithm. This graph-based regrouping suppresses forged correlations because the community structure derives from relational topology rather than raw feature values that adversaries can easily perturb. The resulting community labels then drive both flow chart generation, which extracts decision paths from the gradient boosted tree classifier for representative transactions, and plain-language rule generation, which uses logistic regression to identify the most discriminative features within each community. Furthermore, we introduce a decoupled Kullback-Leibler divergence loss during classifier retraining that penalizes the model for concentrating predictive confidence within a single community, thereby forcing a more uniform distribution of uncertainty across groups. This regularization prevents adversaries from crafting perturbations that collapse all fraudulent transactions into one misleadingly simple cluster. Our framework therefore provides a robust visualization interface that maintains interpretability under adversarial conditions, with the community detection step acting as a structural filter against feature-space manipulation. The proposed method represents a significant departure from static explanation systems, offering a dynamic and adversarially resilient approach to fraud analytics visualization.
References
M Ahmed, AN Mahmood, and MR Islam. A survey of anomaly detection techniques in financial domain. Future Generation Computer Systems, 2016.
Y Chabchoub, MU Togbe, A Boly, and R Chiky. An in-depth study and improvement of isolation forest. IEEE Access, 2022.
B Schölkopf, JC Platt, J Shawe-Taylor, et al. Estimating the support of a high-dimensional distribution. Neural Computation, 2001.
C Zhou and RC Paffenroth. Anomaly detection with robust deep autoencoders. In Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2017.
A Graves. Long short-term memory. Supervised Sequence Labelling with Recurrent Neural Networks, 2012.
MT Ribeiro, S Singh, and C Guestrin. " why should i trust you?" explaining the predictions of any classifier. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations, 2016.
SM Lundberg and SI Lee. A unified approach to interpreting model predictions. In Advances in Neural Information Processing Systems, 2017.
IJ Goodfellow, J Shlens, and C Szegedy. Explaining and harnessing adversarial examples. arxiv, available online. 2014, 2014.
QM Carlo. Monte carlo methods in financial engineering. Technical report, maphysto.dk, 2001.
M Gramopadhye, S Singh, K Agarwal, et al. Curl: Coupled representation learning of cards and merchants to detect transaction frauds. Lecture Notes in Computer Science, 2021.
VD Blondel, JL Guillaume, R Lambiotte, et al. Fast unfolding of communities in large networks. Journal of Statistical Mechanics: Theory and Experiment, 2008.
J Ellson, E Gansner, L Koutsofios, SC North, et al. Graphviz—open source graph drawing tools. In International Symposium on Graph Drawing, 2001.
I Sutskever, O Vinyals, and QV Le. Sequence to sequence learning with neural networks. In Advances in Neural Information Processing Systems, 2014.
C Doersch. Tutorial on variational autoencoders. Technical report, arXiv preprint arXiv:1606.05908, 2016.
G Ridgeway. Generalized boosted models: A guide to the gbm package. Update, 2007.
M Schonlau. Random forests. Applied statistical learning: with case studies in Stata, 2023.
A Olushola and J Mart. Fraud detection using machine learning. ScienceOpen Preprints, 2024.
L Jiang, R Ryan, Q Li, et al. A survey of heterogeneous graph neural networks for cybersecurity anomaly detection. Journal of Computer Security, 2025.
J Amebleh, E Igba, and OM Ijiga. Graph-based fraud detection in open-loop gift cards: Heterogeneous gnns, streaming feature stores, and near-zero-lag anomaly alerts. International Journal of Scientific Research in Science, Engineering and Technology, 2021.
U Zafar and F Wu. Methodological challenges in explainable ai for fraud detection: a systematic literature review. Artificial Intelligence Review, 2026.
S Gee. Fraud and fraud detection: A data analytics approach. Technical report, books.google.com, 2014.
J Vadillo, R Santana, and JA Lozano. Adversarial attacks in explainable machine learning: A survey of threats against models and humans. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 2025.
A Abomakhelb, KA Jalil, AG Buja, A Alhammadi, et al. A comprehensive review of adversarial attacks and defense strategies in deep neural networks. Technologies, 2025.
J Cui, Z Tian, Z Zhong, X Qi, B Yu, and H Zhang. Decoupled kullback-leibler divergence loss. Technical report, arXiv preprint arXiv:2305.13948, 2023.
V Braverman, A Hassidim, Y Matias, et al. Adversarial robustness of streaming algorithms through importance sampling. In Advances in Neural Information Processing Systems, 2021.
S Rakhmetulayeva, A Kulbayeva, et al. Identifying the graphbased typology features for machine learning models in financial fraud detection. Eastern-European Journal of Enterprise Technologies, 2025.
S Gajula. Two pillars of banking intelligence: A comparative analysis of ai techniques for fraud prevention and churn mitigation. In 2026 14th International Symposium on Digital Forensics and Security (ISDFS), 2026.
I Mulyana and MB Ulum. Application of machine learning models for fraud detection in synthetic mobile financial transactions. JITK, 2025.
A Bifet. Adaptive stream mining: Pattern learning and mining from evolving data streams. Technical report, books.google.com, 2010.
D Greene, D Doyle, et al. Tracking the evolution of communities in dynamic social networks. In 2010 International Conference on Advances in Social Networks Analysis and Mining, 2010.
N Turner Lee. Detecting racial bias in algorithms and machine learning. Journal of Information, Communication and Ethics in Society, 2018.
RK Mothilal, A Sharma, and C Tan. Explaining machine learning classifiers through diverse counterfactual explanations. In Proceedings of, 2020.
I Chien, CY Lin, and IH Wang. Community detection in hypergraphs: Optimal statistical limit and efficient algorithms. In International Conference on Machine Learning, 2018.
S Yang, X Zheng, J Li, J Xu, X Wang, et al. Recda: Concept drift adaptation with representation enhancement for network intrusion detection. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2024.
L Gao, H Wang, Z Zhang, H Zhuang, and B Zhou. Hetinf: social influence prediction with heterogeneous graph neural network. Frontiers in Physics, 2022.
NK Vayyasi. Retail fraud analytics using generative intelligence and java cloud frameworks. International Journal of Science, Research and Technology, 2023.
J Ren, F Xia, I Lee, A Noori Hoshyar, et al. Graph learning for anomaly analytics: Algorithms, applications, and challenges. ACM Transactions on Intelligent Systems and Technology, 2023.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 The Author(s)
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
