HYBRID TEMPORAL-GRAPH MACHINE LEARNING FOR ANOMALY DETECTION IN METAVERSE FINANCIAL TRANSACTIONS
Keywords:
Metaverse transactions, hybrid anomaly detection, Graph Neural Networks, temporal modeling, XGBoost, explainable AI (SHAP).Abstract
Metaverse financial ecosystems generate large-scale, pseudonymous, and rapidly evolving transaction networks, creating significant challenges for fraud detection. Traditional rule-based approaches are insufficient to capture the complex structural, temporal, and behavioral patterns present in these systems. This study proposes a Hybrid Temporal-Graph Anomaly Detection (HTGAD) framework that combines three complementary models: a tabular XGBoost classifier for transaction-level features, a GraphSAGE-MLP model for relational patterns in sender-receiver networks, and an LSTM model for sequential behavioral analysis. These outputs are integrated using a logistic regression meta-layer to produce a unified risk score, while SHAP explainability enhances interpretability. An empirical evaluation on the Metaverse Financial Transactions Dataset (78,600 records sourced from Kaggle) shows that XGBoost alone achieves near-perfect classification, LSTM captures temporal dependencies effectively, and GraphSAGE provides structural insights despite its limited standalone performance. The HTGAD fusion model outperforms all individual components, achieving an AUC-ROC of 1.000, an F1-score of 0.9998, a precision of 0.9996, and a recall of 1.000 on the test set. Feature importance analysis identifies transaction type, transaction amount, temporal activity, and user behavior as the most influential predictors of anomalies. These results demonstrate that integrating tabular, temporal, and graph-based perspectives produces a robust, accurate, and interpretable system for detecting anomalous and potentially fraudulent activity in complex, decentralized Metaverse financial environments.
References
Achuthan, K., Ramanathan, S., & Raman, R. (2025). Securing the metaverse: Machine learning-based perspectives on risk, trust, and governance. International Journal of Information Management Data Insights, 5(2), 100356. https://doi.org/10.1016/j.jjimei.2025.100356
Airlangga, G. (2024). Deep learning for anomaly detection and fraud analysis in blockchain transactions of the open metaverse. Jurnal Informatika Ekonomi Bisnis, 6(2). https://doi.org/10.37034/infeb.v6i2.865
Aysan, A. F., Gozgor, G., & Nanaeva, Z. (2024). Technological perspectives of metaverse for financial service providers. Technological Forecasting and Social Change, 202, 123323.
Chen, S., Liu, Y., Zhang, Q., Shao, Z., & Wang, Z. (2025). Multi-distance spatial-temporal graph neural network for anomaly detection in blockchain transactions. Advanced Intelligent Systems, 7(8), 2400898.
Cholevas, C., Angeli, E., Sereti, Z., Mavrikos, E., & Tsekouras, G. E. (2024). Anomaly detection in blockchain networks using unsupervised learning: A survey. Algorithms, 17(5), 201.
El Emary, I. M. M. (2026). Anomaly detection in blockchain-based metaverse transactions using hybrid autoencoder and isolation forest models for risk identification and behavioral pattern analysis. International Journal Research on Metaverse, 3(1), 46–63.
Ferretti, S., D'Angelo, G., & Ghini, V. (2025). Enhancing anti-money laundering frameworks: An application of graph neural networks in cryptocurrency transaction classification. IEEE Access, 13, 50201–50215.
Huynh-The, T., et al. (2023). Blockchain for the metaverse: A review. Future Generation Computer Systems, 143, 401–419.
Iftekhar Rasul, S. M., et al. (2024). Detecting financial fraud in real-time transactions using graph neural networks and anomaly detection. Journal of Economics, Finance and Accounting Studies, 6(1), 131–142.
Jumani, F., & Raza, M. (2025). Machine learning for anomaly detection in blockchain: A critical analysis, empirical validation, and future outlook. Computers, 14(7), 247.
Kim, Y., Lee, Y., Choe, M., Oh, S., & Lee, Y. (2024). Temporal graph networks for graph anomaly detection in financial networks. arXiv. https://doi.org/10.48550/ARXIV.2404.00060
Li, J., Lu, G., Gao, Y., & Gao, F. (2023). A smart contract vulnerability detection method based on multimodal feature fusion and deep learning. Mathematics, 11(23), 4823.
Li, M. (2024). Meta-universe financial transaction anomaly detection and risk prediction based on machine learning. In Proceedings of the 2024 2nd International Conference on Image, Algorithms and Artificial Intelligence (ICIAAI 2024) (Vol. 115, pp. 117–129).
Liu, T., et al. (2024). The role of transformer models in advancing blockchain technology: A systematic survey. arXiv. https://doi.org/10.48550/ARXIV.2409.02139
Macak, M., et al. (2022). Process mining usage in cybersecurity and software reliability analysis: A systematic literature review. Array, 13, 100120.
Mounnan, O., et al. (2024). A review on deep anomaly detection in blockchain. Blockchain: Research and Applications, 5(4), 100227.
Pasaribu, J., Yudistira, N., & Mahmudy, W. F. (2024). Tabular data classification and regression: XGBoost or deep learning with retrieval-augmented generation. IEEE Access, 1–1.
Pocher, N., et al. (2023). Detecting anomalous cryptocurrency transactions: An AML/CFT application of machine learning-based forensics. Electronic Markets, 33(1), 37.
Prakash, S. (2026). Hybrid ensemble learning for anomaly detection in metaverse transactions using isolation forest, autoencoder, and XGBoost. International Journal Research on Metaverse, 3(1), 64–80.
Ranjbaran, G., et al. (2025). C-SHAP: A hybrid method for fast and efficient interpretability. Applied Sciences, 15(2), 672.
Sebt, M. V., Ghasemi, S. H., & Mehrkian, S. S. (2021). Predicting the number of customer transactions using stacked LSTM recurrent neural networks. Social Network Analysis and Mining, 11(1), 86.
Siddamsetti, S. (2024). Anomaly detection in blockchain using machine learning. Journal of Electrical Systems, 20(3), 619–634.
Umer, Q., et al. (2023). Ensemble deep learning-based prediction of fraudulent cryptocurrency transactions. IEEE Access, 11, 95213–95224.
Wang, X., Guo, J., Luo, X., & Yu, H. (2024). DyHDGE: Dynamic heterogeneous transaction graph embedding for safety-centric fraud detection in financial scenarios. Journal of Safety Science and Resilience, 5(4), 486–497.
Xu, B., et al. (2023). Study on the prediction of the uniaxial compressive strength of rock based on the SSA-XGBoost model. Sustainability, 15(6), 5201.
Yang, X., Zhao, X., & Shen, Z. (2024). A generalizable anomaly detection method in dynamic graphs. arXiv. https://doi.org/10.48550/ARXIV.2412.16447
Zhou, L., Zeng, Q., & Li, B. (2022). Hybrid anomaly detection via multihead dynamic graph attention networks for multivariate time series. IEEE Access, 10, 40967–40978.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Jacques Ishimwe, Jonathan Ngugi, Irenee Mungwarakarama
This work is licensed under a Creative Commons Attribution 4.0 International License.
