⚠ Official Notice: www.ijisrt.com is the official website of the International Journal of Innovative Science and Research Technology (IJISRT) Journal for research paper submission and publication. Please beware of fake or duplicate websites using the IJISRT name.



A Hybrid LSTM–Transformer Architecture with Bayesian Uncertainty Quantification for Multi-Horizon Financial Time Series Forecasting


Authors : Bridget Elo Osigho; Charles Amofa; Joy Onma Enyejo; Rukayat Akingbade

Volume/Issue : Volume 11 - 2026, Issue 4 - April

Google Scholar : https://tinyurl.com/h3x5c9mn

Scribd : https://tinyurl.com/472kbj5h

DOI : https://doi.org/10.38124/ijisrt/26apr439

PlumX Metrics

Semantic Scholar

ResearchGate

Note : A published paper may take 4-5 working days from the publication date to appear in PlumX Metrics, Semantic Scholar, and ResearchGate.

Abstract : This study introduces a novel Bayesian Temporal Fusion Transformer-LSTM Hybrid (B-TFTL) for multi-horizon financial time series forecasting, combining the sequence modeling strengths of LSTM with the contextual attention capabilities of Transformer architectures, augmented by Bayesian uncertainty quantification. The proposed algorithm integrates a probabilistic inference layer that captures predictive uncertainty through variational Bayesian techniques, enabling robust forecasting under noisy and non-stationary market conditions. Unlike conventional deterministic models, B-TFTL produces both point forecasts and confidence intervals, improving decision-making in risk-sensitive financial applications. The model is benchmarked against six widely used approaches: Autoregressive Integrated Moving Average (ARIMA), Generalized Autoregressive Conditional Heteroskedasticity (GARCH), standard LSTM, Gated Recurrent Unit (GRU), vanilla Transformer, and Temporal Fusion Transformer (TFT). Experimental results across equity indices, forex, and commodity datasets show that B-TFTL consistently outperforms these models in terms of lower root mean squared error (RMSE), improved directional accuracy, and better calibration of predictive intervals. The hybrid architecture effectively captures both short-term dependencies and long-range temporal patterns, while the Bayesian component enhances robustness to volatility clustering and structural shifts. Additionally, attention weight analysis provides interpretability by identifying key temporal features influencing forecasts. The proposed framework advances financial forecasting by unifying deep learning and probabilistic modeling, offering a powerful and reliable tool for multi-horizon prediction in complex financial environments.

Keywords : Bayesian Temporal Fusion Transformer-LSTM Hybrid (B-TFTL); Multi-Horizon Forecasting; Financial Time Series; Uncertainty Quantification; Attention Mechanisms.

References :

  1. Ajayi, A. A., Igba, E., Soyele, A. D., & Enyejo, J. O. (2024). Enhancing digital identity and financial security in decentralized finance through zero-knowledge proofs and blockchain solutions. IRE Journals, 8(4).
  2. Akorli, K. Y., & Enyejo, J. O. (2024). Developing causal uplift algorithm for US omnichannel personalization optimizing lifetime value predictions. International Journal of Scientific Research in Computer Science, Engineering and Information Technology, 10(6), 2603–2623. https://doi.org/10.32628/CSEIT25113677
  3. Aluso, L., & Enyejo, J. O. (2024). Leveraging NLP and retrieval-augmented generation (RAG) models for automated business intelligence query resolution. International Journal of Scientific Research in Science, Engineering and Technology, 11(4), 534–557. https://doi.org/10.32628/IJSRSET242439
  4. Amofa, C., Osigho, B. E., Enyejo, J. O., & Akingbade, R. (2025). Graph Neural Network-Based Cross-Asset Pricing Model with Adaptive Factor Learning and Temporal Attention for High-Dimensional Financial Markets. International Journal of Scientific Research and Modern Technology, 4(11), 207–223. https://doi.org/10.38124/ijsrmt.v4i11.1353
  5. Animasaun, J. B., Ijiga, O. M., Ayoola, V. B., & Enyejo, L. A. (2024). Impact of solvent polarity on cannabinoid recovery: A multivariate optimization study. International Journal of Scientific Research and Modern Technology, 3(1), 40–54.
  6. Anokwuru, E. A. (2024). Leveraging AI-enhanced commercial insights for precision marketing. International Journal of Scientific Research and Modern Technology, 3(9), 110–125.
  7. Tom-Ayegunle, K., et al. (2025). Cumulative burden of geriatric conditions and cardiovascular outcomes. JACC Advances, 4(12).
  8. Armah, G. D., Idoko, P. I., Adeyeye, Y. I., Enyejo, L. A., & Azonuche, T. I. (2024). Quantifying the economic spillover effects of healthcare data breaches using panel regression. European Journal of Biomedical and Pharmaceutical Sciences, 11(12), 631–656.
  9. Ashish, V. (2017). Attention is all you need. Advances in neural information processing systems, 30, I.
  10. Bianchi, D., Büchner, M., & Tamoni, A. (2021). Bond risk premia with machine learning. The Review of Financial Studies, 34(2), 1046–1089.
  11. Blundell, C., Cornebise, J., Kavukcuoglu, K., & Wierstra, D. (2015). Weight uncertainty in neural network. In International conference on machine learning (pp. 1613-1622). PMLR.
  12. Cho, K., et al. (2014). Learning phrase representations using RNN encoder–decoder for statistical machine translation. EMNLP.
  13. Dankwah, H. A. K., & Enyejo, J. O. (2024). Proposing JournalGuard a graph neural network algorithm for continuous audit of ERP journal entries with explainable control-risk scoring. International Journal of Scientific Research and Modern Technology, 3(7), 66–83. https://doi.org/10.38124/ijsrmt.v3i7.1296
  14. Dankwah, H. A. K., & Enyejo, J. O. (2025). AI-driven explainable predictive analytics framework for fair value measurement and earnings quality prediction. International Journal of Scientific Research and Modern Technology, 4(6), 110–126. https://doi.org/10.38124/ijsrmt.v4i6.1315
  15. Engle, R. F. (2016). Dynamic conditional beta. Journal of Financial Econometrics, 14(4), 643–667.
  16. Enyejo, J. O., et al. (2024). Data-driven digital marketing and battery supply chain optimization. International Journal of Scholarly Research and Reviews, 5(2), 1–20.
  17. Fischer, T., & Krauss, C. (2018). Deep learning with long short-term memory networks for financial market predictions. European Journal of Operational Research, 270(2), 654–669.
  18. Gal, Y., & Ghahramani, Z. (2015). Dropout as a Bayesian approximation. Proceedings of ICML.
  19. Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation.
  20. Hyndman, R. J., & Athanasopoulos, G. (2018). Forecasting: Principles and practice. OTexts.
  21. Kendall, A., & Gal, Y. (2017). What uncertainties do we need in Bayesian deep learning for computer vision? Advances in Neural Information Processing Systems.
  22. Kwarteng, R. A., Idoko, I. P., & Ijiga, O. M. (2021). Data-driven project management frameworks for improving IT service delivery in distributed organizations. Computer Science & IT Research Journal, 2(1).
  23. Lim, B., Arık, S. Ö., Loeff, N., & Pfister, T. (2021). Temporal fusion transformers for interpretable multi-horizon time series forecasting. International Journal of Forecasting, 37(4), 1748–1764.
  24. Ogbuonyalu, U. O., Abiodun, K., Dzamefe, S., Vera, E. N., Oyinlola, A., & Igba, E. (2024). Assessing artificial intelligence driven algorithmic trading implications on market liquidity risk and financial systemic vulnerabilities. International Journal of Scientific Research and Modern Technology, 3(4), 18–21. https://doi.org/10.38124/ijsrmt.v3i4.433
  25. Ononiwu, M., Azonuche, T. I., & Enyejo, J. O. (2023). Exploring influencer marketing using encrypted CRM analytics and adaptive progressive web app development. International Journal of Scientific Research and Modern Technology, 2(6), 1–13.
  26. Onwuzurike, M. A., Enyejo, J. O., & Peter-Anyebe, A. C. (2026). Design and evaluation of real-time adaptive learning algorithms. World Journal of Advance Multidisciplinary Research, 3(3), 21–36.
  27. Rasul, K., Sheikh, A. S., Schuster, I., Bergmann, U., & Vollgraf, R. (2020). Multivariate probabilistic time series forecasting via conditioned normalizing flows. arXiv preprint arXiv:2002.06103.
  28. Sanmori, M. T. (2024). AI-driven functional independence prediction and assistive technology optimization. International Journal of Scientific Research and Modern Technology, 3(11), 186–205.
  29. Sanmori, M. T. (2024). AI-driven functional independence prediction and assistive technology optimization. International Journal of Scientific Research and Modern Technology, 3(11), 186–205.
  30. Sezer, O. B., Gudelek, M. U., & Ozbayoglu, A. M. (2020). Financial time series forecasting with deep learning: A systematic literature review. Applied Soft Computing, 90, 106181.
  31. Usoro, S. O., & Amunigun, A. A. (2024). Public–private partnerships in strengthening rural food supply chains. International Journal of Scientific Research in Science, Engineering and Technology, 11(2), 645–659.
  32. Usoro, S. O., Galadima, E. R., & Adogwa, O. H. (2025). Cold chain logistics optimization integrating IoT and data analytics. International Journal of Scientific Research in Science & Technology, 12(2), 1452–1468.
  33. Zhang, Y., Aggarwal, C., & Qi, G. (2017). Stock price prediction via discovering multi-frequency trading patterns. KDD Conference Proceedings.
  34. Zhang, Z., Zohren, S., & Roberts, S. (2020). Deep learning for portfolio optimization. Journal of Financial Data Science, 2(4), 8–20.

This study introduces a novel Bayesian Temporal Fusion Transformer-LSTM Hybrid (B-TFTL) for multi-horizon financial time series forecasting, combining the sequence modeling strengths of LSTM with the contextual attention capabilities of Transformer architectures, augmented by Bayesian uncertainty quantification. The proposed algorithm integrates a probabilistic inference layer that captures predictive uncertainty through variational Bayesian techniques, enabling robust forecasting under noisy and non-stationary market conditions. Unlike conventional deterministic models, B-TFTL produces both point forecasts and confidence intervals, improving decision-making in risk-sensitive financial applications. The model is benchmarked against six widely used approaches: Autoregressive Integrated Moving Average (ARIMA), Generalized Autoregressive Conditional Heteroskedasticity (GARCH), standard LSTM, Gated Recurrent Unit (GRU), vanilla Transformer, and Temporal Fusion Transformer (TFT). Experimental results across equity indices, forex, and commodity datasets show that B-TFTL consistently outperforms these models in terms of lower root mean squared error (RMSE), improved directional accuracy, and better calibration of predictive intervals. The hybrid architecture effectively captures both short-term dependencies and long-range temporal patterns, while the Bayesian component enhances robustness to volatility clustering and structural shifts. Additionally, attention weight analysis provides interpretability by identifying key temporal features influencing forecasts. The proposed framework advances financial forecasting by unifying deep learning and probabilistic modeling, offering a powerful and reliable tool for multi-horizon prediction in complex financial environments.

Keywords : Bayesian Temporal Fusion Transformer-LSTM Hybrid (B-TFTL); Multi-Horizon Forecasting; Financial Time Series; Uncertainty Quantification; Attention Mechanisms.

Paper Submission Last Date
30 - April - 2026

SUBMIT YOUR PAPER CALL FOR PAPERS
Video Explanation for Published paper

Never miss an update from Papermashup

Get notified about the latest tutorials and downloads.

Subscribe by Email

Get alerts directly into your inbox after each post and stay updated.
Subscribe
OR

Subscribe by RSS

Add our RSS to your feedreader to get regular updates from us.
Subscribe