Foundation Models for Financial Market Stress Forecasting: A Leakage-Safe Benchmark of Time-Series Transformers, Classical Machine Learning, and Explainable Risk Indicators
Keywords:
financial stress forecasting, leakage-safe benchmark, time-series transformers, classical machine learning, explainable risk indicators, systemic risk, walk-forward validation, model governanceAbstract
Financial market stress forecasting has become a critical component of systemic risk management, yet the methodological integrity of many predictive models is compromised by data leakage, particularly in benchmark evaluations. This paper presents a comprehensive, leakage-safe benchmark for comparing foundation-scale time-series transformers, classical machine learning algorithms, and explainable risk indicators in the context of early-warning market stress detection. We introduce a novel evaluation framework that strictly enforces temporal walk-forward validation, feature independence from future information, and label construction that avoids look-ahead bias. The benchmark covers multiple asset classes and stress regimes, including volatility spikes, drawdown events, and correlation breakdowns. Our results demonstrate that while transformer-based models achieve superior predictive accuracy on raw time-series data, their advantage diminishes significantly under leakage-safe conditions, where classical tree ensembles and parsimonious risk indicators remain competitive. We further examine the trade-offs between model complexity, interpretability, and deployment sustainability, arguing that system-level robustness requires not only high predictive performance but also governance-aware design that prevents feedback loops and fairness violations. The study concludes with policy recommendations for regulatory adoption of leakage-safe evaluation standards and the integration of explainable indicators into stress testing infrastructure. Our findings underscore the necessity of rigorous benchmark design to ensure that machine learning advances translate into credible and actionable tools for financial stability.
References
1. Liu, T. (2026). Leakage-Safe Benchmark Design for Market-Stress Early Warning: An Economically Credible Evaluation.
2. Hu, L., & Shen, Y. (2026). A predictive analytics approach for forecasting global stock index returns using deep learning techniques. Decision Analytics Journal, 100685.
3. 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 (pp. 785–794). ACM.
4. Liu, T. (2026). Beyond volatility: A leakage-safe residual-stress signal for drawdown risk monitoring. Available at SSRN 6503179.
5. Liu, T. (2026). PCA-APT Stress Index for Market Drawdowns.
6. Bollerslev, T. (1986). Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics, 31(3), 307–327.
7. Alizadeh, S., Brandt, M. W., & Diebold, F. X. (2002). Range-based estimation of stochastic volatility models. Journal of Finance, 57(3), 1047–1091.
8. Bergmeir, C., & Benítez, J. M. (2012). On the use of cross-validation for time series prediction. Information Sciences, 191, 192–213.
9. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017). Attention is all you need. In Advances in Neural Information Processing Systems 30 (pp. 5998–6008).
10. Wu, H., Xu, J., Wang, J., & Long, M. (2021). Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting. In Advances in Neural Information Processing Systems 34 (pp. 22419–22430).
11. Nesvetailova, A., & Palan, R. (2020). Sabotage in the financial system: Lessons from the last crisis. Journal of Economic Issues, 54(2), 457–464.
12. Nie, Y., Nguyen, N. H., Sinthong, P., & Kalagnanam, J. (2023). A time series is worth 64 words: Long-term forecasting with transformers. In International Conference on Learning Representations (ICLR).
13. Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2018). Statistical and machine learning forecasting methods: Concerns and ways forward. PLoS ONE, 13(3), e0194889.
14. Board of Governors of the Federal Reserve System. (2021). Comprehensive Capital Analysis and Review 2021: Summary Instructions. Federal Reserve.
15. European Banking Authority. (2021). Guidelines on model risk management. EBA/GL/2021/04.
16. Strubell, E., Ganesh, A., & McCallum, A. (2019). Energy and policy considerations for deep learning in NLP. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (pp. 3645–3650).
17. Dwork, C., Hardt, M., Pitassi, T., Reingold, O., & Zemel, R. (2012). Fairness through awareness. In Proceedings of the 3rd Innovations in Theoretical Computer Science Conference (pp. 214–226).
18. Liu, T. (2022, December). Financial Constraint’Impact on Firms’ ESG Rating Based on Chinese Stock Market. In 2022 4th International Conference on Economic Management and Cultural Industry (ICEMCI 2022) (pp. 1085-1095). Atlantis Press.
19. Liu, T. (2026). Interpretable Machine Learning for Volatility Forecasting Under Realistic Walk-Forward Constraints.
20. Liu, T. (2026). A Comparative Study of Transformer-Based and Classical Models for Financial Time-Series Forecasting. Journal of Risk and Financial Management, 19(3), 203.
21. Liu, T. (2026). Volatility Forecasting and Early-Warning Market Stress Detection: A Leakage-Safe Evaluation with Tree Ensembles and Transformers.
22. Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., & Liu, T.-Y. (2017). LightGBM: A highly efficient gradient boosting decision tree. In Advances in Neural Information Processing Systems 30 (pp. 3146–3154).
23. Zhang, G. P. (2003). Time series forecasting using a hybrid ARIMA and neural network model. Neurocomputing, 50, 159–175.
24. Hansen, L. P., & Sargent, T. J. (2008). Robustness. Princeton University Press.
25. Athey, S., & Imbens, G. W. (2016). Recursive partitioning for heterogeneous causal effects. Proceedings of the National Academy of Sciences, 113(27), 7353–7360.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Global Financial Analytics Research Review
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.
