Comparative Performance of Temporal Fusion Transformer for Multi-horizon Price Forecasting and Silver Risk Assessment for Thai Importers

Authors

  • Natee Suwanwetin Independent Researcher, Bangkok, Thailand

Keywords:

Temporal Fusion Transformer (TFT), multi-horizon forecasting, risk assessment, silver, financial engineering

Abstract

This research aims to develop and compare the performance of the Temporal Fusion Transformer (TFT) model for multi-horizon price forecasting and silver risk assessment for importers in Thailand. A daily time-series dataset of 2,450 observations, spanning from 2015 to 2025, was utilized. The data was partitioned into training, validation, and testing sets with a ratio of 80:10:10, respectively. The evaluation of model performance covered forecasting horizons of 1, 7, and 30 days to assess predictive accuracy across both short-term and medium-term periods.

The empirical results reveal that the TFT model significantly outperforms traditional benchmarks. Specifically, for the 1-day forecasting horizon, the TFT achieved an RMSE of 0.1673 and a MAPE of 0.5014%, demonstrating substantially higher accuracy than the LSTM (RMSE = 1.12) and GARCH (RMSE = 1.85) models. Furthermore, the quantile forecasting mechanism within the TFT architecture effectively estimated Value-at-Risk (VaR) at 95% and 99% confidence levels, with results consistently aligned with backtesting standards under the Kupiec Test.

A key finding indicates that the USD/THB exchange rate is the most influential factor on forecasting accuracy, accounting for 42% of the relative importance. Consequently, the results of this study serve as a vital strategic decision-support tool for practitioners in formulating hedging strategies and managing import costs effectively amidst high market volatility.

References

Aparicio, F., Morales, A. J., & Guerrero, R. (2022). Precious metals and macroeconomic factors: A deep learning approach for multi-step ahead forecasting. Resources Policy, 78, 102812. https://doi.org/10.1016/j.resourpol.2022.102812

Arratia, A., & Lopez-Barrado, A. (2021). Evaluation of value-at-risk models using deep learning: Evidence from precious metal markets. Journal of Risk and Financial Management, 14(11), 541. https://doi.org/10.3390/jrfm14110541

Baur, D. G., & McDermott, T. K. (2010). Is gold a safe haven? International evidence. Journal of Banking & Finance, 34(8), 1886-1898. https://doi.org/10.1016/j.jbankfin.2009.12.008

Buehler, H., Gonon, L., Teichmann, J., & Wood, B. (2019). Deep hedging. Quantitative Finance, 19(8), 1271-1291. https://doi.org/10.1080/14697688.2019.1571683

Chadsuthi, S. (2021). Forecasting gold prices in Thailand using ARIMA and machine learning models. Journal of Science and Technology, 29(4), 645-658.

Chen, W., Zhang, H., & Mehlawat, M. K. (2022). Deep learning-based value-at-risk forecasting for commodity markets. International Review of Financial Analysis, 81, 102087. https://doi.org/10.1016/j.irfa.2022.102087

Engle, R. F. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica, 50(4), 987-1007. https://doi.org/10.2307/1912773

Hamid, S. A., & Al-Ghazali, M. (2022). Attention mechanisms in transformer-based models for commodity price forecasting. Expert Systems with Applications, 191, 116245. https://doi.org/10.1016/j.eswa.2021.116245

Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735

Jareño, F., de la O González, M., & Tolentino, M. (2021). Precious metals and the US dollar index: A dynamic relationship analysis. Resources Policy, 70, 101905. https://doi.org/10.1016/j.resourpol.2020.101905

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. https://doi.org/10.1016/j.ijforecast.2021.03.012

Park, J., Lee, K., & Kim, H. (2024). Risk management in commodity markets using attention-based deep learning models. Journal of Commodity Markets, 33, 100389. https://doi.org/10.1016/j.jcomm.2023.100389

Smith, T., & Johnson, L. (2023). The shifting correlation between gold and silver in the post-pandemic era. Financial Markets and Portfolio Management, 37(2), 145-168. https://doi.org/10.1007/s11408-023-00421-w

Tiwari, A. K., Abakah, E. J. A., & Gabauer, D. (2023). Exchange rate volatility and its impact on commodity prices in emerging markets. Economic Modelling, 118, 106093. https://doi.org/10.1016/j.econmod.2022.106093

Wu, N., Green, B., Ben, X., & O'Banion, S. (2020). Deep transformer models for time series forecasting: The influenza prevalence case. arXiv. https://doi.org/10.48550/arXiv.2001.08317

Downloads

Published

2026-04-17

How to Cite

Suwanwetin, N. (2026). Comparative Performance of Temporal Fusion Transformer for Multi-horizon Price Forecasting and Silver Risk Assessment for Thai Importers. Siam University Journal of Business Administration, 27(48), 133–147. retrieved from https://so07.tci-thaijo.org/index.php/sujba/article/view/10309

Issue

Section

Research Articles