JDS Journal of Data Science 1683-86021680-743X1680-743X School of Statistics, Renmin University of China JDS1190 10.6339/25-JDS1190 Data Science in Action Money Laundering Detection with Multi-Aggregation Custom Edge GIN https://orcid.org/0000-0001-5938-7260 WójcikFilipfilip.wojcik@ue.wroc.plds@filip-wojcik.com1∗ Faculty of Business Intelligence, Komandorska 118/120 54-132 Wroclaw, Wroclaw University of Economics and Business, Poland Email: filip.wojcik@ue.wroc.pl or ds@filip-wojcik.com. 2025126202500119 Supplementary Material

The source code for this study is available on GitHub: https://github.com/maddataanalyst/Graph_MAGIC_Conv. The repository includes all the necessary components to reproduce the training results.

A supplementary PDF file attached to this publication provides detailed analyses of the train/validation/test splits, hyperparameter tuning results, and a comprehensive breakdown of the model architecture for each dataset.

30920242352025 2025 The Author(s). Published by the School of Statistics and the Center for Applied Statistics, Renmin University of China.2025 Open access article under the CC BY license.

Detecting illicit transactions in Anti-Money Laundering (AML) systems remains a significant challenge due to class imbalances and the complexity of financial networks. This study introduces the Multiple Aggregations for Graph Isomorphism Network with Custom Edges (MAGIC) convolution, an enhancement of the Graph Isomorphism Network (GIN) designed to improve the detection of illicit transactions in AML systems. MAGIC integrates edge convolution (GINE Conv) and multiple learnable aggregations, allowing for varied embedding sizes and increased generalization capabilities. Experiments were conducted using synthetic datasets, which simulate real-world transactions, following the experimental setup of previous studies to ensure comparability. MAGIC, when combined with XGBoost as a link predictor, outperformed existing models in 16 out of 24 metrics, with notable improvements in F1 scores and precision. In the most imbalanced dataset, MAGIC achieved an F1 score of 82.6% and a precision of 90.4% for the illicit class. While MAGIC demonstrated high precision, its recall was lower or comparable to the other models, indicating potential areas for future enhancement. Overall, MAGIC presents a robust approach to AML detection, particularly in scenarios where precision and overall quality are critical. Future research should focus on optimizing the model’s recall, potentially by incorporating additional regularization techniques or advanced sampling methods. Additionally, exploring the integration of foundation models like GraphAny could further enhance the model’s applicability in diverse AML environments.

 deep learning financial fraud detection graph neural networks graph representation learning References Alarab I, Prakoonwit S, Nacer MI (2020). Competence of graph convolutional networks for anti-money laundering in bitcoin blockchain. In: ICMLT 2020: 5th International Conference on Machine Learning Technologies. Beijing, China, June, 2020, 23–27. Association for Computing Machinery. Alsuwailem AAS, Saudagar AKJ (2020). Anti-money laundering systems: A systematic literature review. Journal of Money Laundering Control, 23: 833–848. https://doi.org/10.1108/JMLC-02-2020-0018 Behera DK, Das M, Swetanisha S, Nayak J, Vimal S, Naik B (2021). Follower link prediction using the xgboost classification model with multiple graph features. Wireless Personal Communications, 127(1): 695–714. https://doi.org/10.1007/s11277-021-08399-y Bouthillier X, Delaunay P, Bronzi M, Trofimov A, Nichyporuk B, Szeto J, et al. (2021). Accounting for variance in machine learning benchmarks. In: Proceedings of Machine Learning and Systems (A Smola, A Dimakis, I Stoica, eds.), volume 3, 747–769. Brossard R, Frigo O, Dehaene D (2020). Graph convolutions that can finally model local structure. CoRR. arXiv preprint: https://arxiv.org/abs/2011.15069. Bruna J, Zaremba W, Szlam A, LeCun Y (2014). Spectral networks and deep locally connected networks on graphs. In: Conf. on Learning Representations, ICLR 2014—Conference Track Proceedings (Y Bengio, Y LeCun, eds.), 1–14. Chen T, Guestrin C (2016). Xgboost: A scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining—KDD ’16, 785–794. ACM Press. Corso G, Cavalleri L, Beaini D, Liò P, Velickovic P (2020). Principal neighbourhood aggregation for graph nets. Advances in Neural Information Processing Systems, 33: 13260–13271. Defferrard M, Bresson X, Vandergheynst P (2016). Convolutional neural networks on graphs with fast localized spectral filtering. Advances in Neural Information Processing Systems, 29: 3837–3845. Dumitrescu B, Baltoiu A, Budulan S (2022). Anomaly detection in graphs of bank transactions for anti money laundering applications. IEEE Access, 10: 47699–47714. https://doi.org/10.1109/ACCESS.2022.3170467 Eddin AN, Bono J, Aparício D, Polido D, Ascensão JT, Bizarro P, et al. (2021). Anti-money laundering alert optimization using machine learning with graphs. CoRR. arXiv preprint: https://arxiv.org/abs/2112.07508. Gallicchio C, Micheli A (2010). Graph echo state networks. In: The 2010 International Joint Conference on Neural Networks (IJCNN), 1–10. IEEE. Gilmer J, Schoenholz SS, Riley PF, Vinyals O, Dahl GE (2017). Neural message passing for quantum chemistry. In: ICML (D Precup, YW Teh, eds.), volume 70 of Proceedings of Machine Learning Research, PMLR, 1263–1272. Gori M, Monfardini G, Scarselli F (2005). A new model for learning in graph domains. In: Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005, 729–734. Hamilton WL (2020). Graph representation learning Hamilton. Synthesis Lectures on Artificial Intelligence and Machine Learning, 14: 1–159. https://doi.org/10.1007/978-3-031-01588-5 Hamilton WL, Ying Z, Leskovec J (2017). Inductive representation learning on large graphs. In: NIPS (I Guyon, U von Luxburg, S Bengio, HM Wallach, R Fergus, SVN Vishwanathan, R Garnett, eds.), 1024–1034. Han J, Huang Y, Liu S, Towey K (2020). Artificial intelligence for anti-money laundering: A review and extension. Digital Finance, 2: 211–239. https://doi.org/10.1007/s42521-020-00023-1 Hu W, Liu B, Gomes J, Zitnik M, Liang P, Pande VS, et al. (2019). Pre-training graph neural networks. CoRR. arXiv preprint: https://arxiv.org/abs/1905.12265. Izmailov P, Podoprikhin D, Garipov T, Vetrov D, Wilson AG (2018). Averaging weights leads to wider optima and better generalization. In: UAI (A Globerson, R Silva, eds.), volume 2, 876–885. AUAI Press. Johannessen F, Jullum M (2023). Finding money launderers using heterogeneous graph neural networks. CoRR. arXiv preprint: https://arxiv.org/abs/2307.13499. Kipf TN, Welling M (2017). Semi-supervised classification with graph convolutional networks. In: ICLR Poster, 1–14. Leskovec J, Faloutsos C (2006). Sampling from large graphs. In: KDD (T Eliassi-Rad, LH Ungar, M Craven, D Gunopulos, eds.), volume 2006, 631–636. Association for Computing Machinery. Leskovec J, Sosič R (2016). Snap: A general-purpose network analysis and graph-mining library. ACM Transactions on Intelligent Systems and Technology, 8: 1–20. Li G, Xiong C, Qian G, Thabet A, Ghanem B (2023). Deepergcn: Training deeper gcns with generalized aggregation functions. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45: 13024–13034. Liu F, Xue S, Wu J, Zhou C, Hu W, Paris C, et al. (2020). Deep learning for community detection: Progress, challenges and opportunities. In: IJCAI, 4981–4987. ijcai.org. Opitz J, Burst S (2019). Macro F1 and macro F1. CoRR. arXiv preprint: https://arxiv.org/abs/1911.03347. Parliament TE (2015). Directive (eu) 2015/ 849 of the european parliamend and of the council—of 20 May 2015. Official Journal of the European Union. Raschka S (2018). Model evaluation, model selection, and algorithm selection in machine learning performance estimation: Generalization performance vs. model selection. arXiv preprint. Sammut C, Webb GI (2011). Encyclopedia of Machine Learning. Springer Science & Business Media. Scarselli F, Gori M, Tsoi AC, Hagenbuchner M, Monfardini G (2009). Computational capabilities of graph neural networks. IEEE Transactions on Neural Networks, 20: 81–102. https://doi.org/10.1109/TNN.2008.2005141 Shi C (2022). Heterogeneous Graph Neural Networks. 351–369. Springer Nature Singapore, Singapore. Silva ÍDG, Correia LHA, Maziero EG (2023). Graph neural networks applied to money laundering detection in intelligent information systems. In: Proceedings of the XIX Brazilian Symposium on Information Systems (MXC da Cunha, MF de Souza Júnior, JC Marques, TM de Classe, RD Araújo, eds.), 252–259. Tailor SA, Opolka FL, Liò P, Lane ND (2022). Do we need anisotropic graph neural networks? arXiv preprint: https://arxiv.org/abs/2104.01481. Thornberry WMM (2021). National defense authorization act for fiscal year 2021. Public Law, 116: 283. Velickovic P, Cucurull G, Casanova A, Romero A, Liò P, Bengio Y (2017). Graph attention networks. CoRR. arXiv preprint: https://arxiv.org/abs/1710.10903. Watanabe S, Bansal A, Hutter F (2023). Ped-anova: Efficiently quantifying hyperparameter importance in arbitrary subspaces. In: Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, IJCAI ’23 (E Elkind, ed.), 4389–4396. Weber M, Chen J, Suzumura T, Pareja A, Ma T, Kanezashi H, et al. (2018). Scalable graph learning for anti-money laundering: A first look. CoRR. arXiv preprint: https://arxiv.org/abs/1812.00076. Weber M, Domeniconi G, Chen J, Weidele DKI, Bellei C, Robinson T, et al. (2019). Anti-money laundering in bitcoin: Experimenting with graph convolutional networks for financial forensics. CoRR. arXiv preprint: https://arxiv.org/abs/1908.02591. Wu Z, Pan S, Chen F, Long G, Zhang C, Yu PS (2021). A comprehensive survey on graph neural networks. IEEE Transactions on Neural Networks and Learning Systems, 32: 4–24. https://doi.org/10.1109/TNNLS.2020.2978386 Xu K, Hu W, Leskovec J, Jegelka S (2018). How powerful are graph neural networks? CoRR. arXiv preprint: https://arxiv.org/abs/1810.00826. Yang C, Xiao Y, Zhang Y, Sun Y, Han J (2020). Heterogeneous network representation learning: A unified framework with survey and benchmark. IEEE Transactions on Knowledge and Data Engineering, 34: 4854–4873. https://doi.org/10.1109/TKDE.2020.3045924 Yang Y, Li D (2020). NENN: Incorporate node and edge features in graph neural networks. In: Proceedings of the 12th Asian Conference on Machine Learning, ACML 2020, 18–20 November 2020, Bangkok, Thailand (SJ Pan, M Sugiyama, eds.), volume 129 of Proceedings of Machine Learning Research, PMLR, 593–608. Ying C, Cai T, Luo S, Zheng S, Ke G, He D, et al. (2021). Do transformers really perform badly for graph representation? Advances in Neural Information Processing Systems, 34: 28877–28888. You J, Ying R, Leskovec J (2020). Design space for graph neural networks. In: Proceedings of the 34th International Conference on Neural Information Processing Systems, NIPS ’20 (H Larochelle, M Ranzato, R Hadsell, M-F Balcan, H-T Lin, eds.), 17009–17021. Curran Associates Inc., Red Hook, NY, USA. Zhao J, Mostafa H, Galkin M, Bronstein MM, Zhu Z, Tang J (2024). Graphany: A foundation model for node classification on any graph. arXiv preprint: https://arxiv.org/abs/2405.20445.