Graph Neural Network-Based Cross-Asset Volatility Propagation with Interpretable Structure Learning Under Realistic Walk-Forward Evaluation
Keywords:
graph neural networks, volatility propagation, structure learning, interpretability, walk-forward evaluation, financial risk managementAbstract
Financial volatility modeling has traditionally relied on univariate or low-dimensional multivariate frameworks that assume static correlation structures and ignore complex, non-linear propagation channels across assets. This paper presents a graph neural network architecture that jointly learns the latent graph topology representing cross-asset volatility linkages and models the directed propagation of volatility shocks through this learned structure. The framework integrates an interpretable structure learning module that produces sparse, causal-adjacent representations of interdependencies, enabling domain experts to audit and validate the learned relationships. To ensure robustness against overfitting and temporal non-stationarity, a realistic walk-forward evaluation protocol is adopted that strictly preserves temporal order, uses expanding windows for training and fixed windows for validation, and accounts for transaction costs and liquidity constraints. The paper discusses the architectural trade-offs between expressivity and transparency, the computational infrastructure required for real-time inference, and the broader socio-technical implications of deploying such models in automated risk management systems. Particular attention is given to fairness and policy concerns arising from unequal access to predictive analytics across market participants, as well as systemic risk externalities introduced when many actors rely on similar graph-based forecasts. By grounding the evaluation in a rigorous out-of-sample regime that mirrors actual trading conditions, the study aims to bridge the gap between academic volatility forecasting and practical financial infrastructure. The findings underscore that while graph neural networks can capture higher-order spillover effects that improve volatility prediction, the interpretability constraints and computational overhead demand careful governance to prevent misuse and ensure market stability.
References
1. Cont, R. (2001). Empirical properties of asset returns: Stylized facts and statistical issues. Quantitative Finance, 1(2), 223-236.
2. Andersen, T. G., Bollerslev, T., & Diebold, F. X. (2007). Roughing it up: Including jump components in the measurement, modeling, and forecasting of return volatility. The Review of Economics and Statistics, 89(4), 701-720.
3. Diebold, F. X., & Yılmaz, K. (2014). On the network topology of variance decompositions: Measuring the connectedness of financial firms. Journal of Econometrics, 182(1), 119-134.
4. Kipf, T. N., & Welling, M. (2017). Semi-supervised classification with graph convolutional networks. In Proceedings of the 5th International Conference on Learning Representations (ICLR).
5. Veličković, P., Cucurull, G., Casanova, A., Romero, A., Lio, P., & Bengio, Y. (2018). Graph attention networks. In Proceedings of the 6th International Conference on Learning Representations (ICLR).
6. Xu, K., Hu, W., Leskovec, J., & Jegelka, S. (2019). How powerful are graph neural networks? In Proceedings of the 7th International Conference on Learning Representations (ICLR).
7. Lopez de Prado, M. (2018). Advances in Financial Machine Learning. John Wiley & Sons.
8. Feng, F., Chen, H., He, X., Ding, J., Sun, M., & Chua, T. S. (2021). Walk-forward analysis for financial time series forecasting. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 502-512.
9. Engle, R. F. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica, 50(4), 987-1007.
10. Bollerslev, T. (1986). Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics, 31(3), 307-327.
11. Hirsa, A., & Nekipelov, D. (2022). Machine Learning in Finance: From Theory to Practice. Cambridge University Press.
12. Carriero, A., Clark, T. E., & Marcellino, M. (2022). Large volatility matrix prediction: A network approach. Journal of Econometrics, 230(2), 370-392.
13. Liu, T. (2026). Interpretable Machine Learning for Volatility Forecasting Under Realistic Walk-Forward Constraints.
14. Monti, F., Boscaini, D., Masci, J., Rodolà, E., Svoboda, J., & Bronstein, M. M. (2017). Geometric deep learning on graphs and manifolds using mixture model CNNs. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 5115-5124.
15. Wu, Z., Pan, S., Chen, F., Long, G., Zhang, C., & Yu, P. S. (2023). A comprehensive survey on graph neural networks. IEEE Transactions on Neural Networks and Learning Systems, 34(4), 2115-2134.
16. Zhang, Z., Cui, P., & Zhu, W. (2023). Deep learning on graphs: A survey. IEEE Transactions on Knowledge and Data Engineering, 35(1), 96-115.
17. Liang, D., Zhang, Y., & Zhou, G. (2024). Dynamic graph learning for financial time series forecasting. In Proceedings of the 38th AAAI Conference on Artificial Intelligence, 8850-8858.
18. Pearl, J. (2019). The seven tools of causal inference, with reflections on machine learning. Communications of the ACM, 62(3), 54-60.
19. Zhu, J., Yan, Y., Zhao, L., & Yu, P. S. (2025). Interpretable graph neural networks: A survey. ACM Computing Surveys, 57(2), 1-38.
20. Li, Z., Zhang, J., & Zhao, H. (2023). Forecasting volatility with graph neural networks. Quantitative Finance, 23(4), 567-582.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2024 Computer Science and Engineering Transactions
This work is licensed under a Creative Commons Attribution 4.0 International License.
This article is published under the Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
