In this paper, we evaluated whether large pretrained time-series models improve financial forecasting across global markets. Using over 10,000 assets spanning equities, bonds, and currencies, we compared ARIMA, XGBoost, LSTM, TimeGPT, and TimesFM under a rolling out-of-sample framework. We found that large pretrained models enhance cross-market stability and consistency, particularly in structurally stable markets such as Japan. However, gains in predictive accuracy were modest and highly conditional: XGBoost remained competitive in bonds and currencies, and no model dominated in efficient markets. Overall, the evidence suggests that pretrained models improve robustness rather than deliver universally superior forecasts, and that statistical improvements do not necessarily translate into economically meaningful advantages.
Citation: Linghan Huang, Yunhan Li, Xuan Feng. Do pretrained models improve financial time series forecasting?[J]. Quantitative Finance and Economics, 2026, 10(4): 728-752. doi: 10.3934/QFE.2026028
In this paper, we evaluated whether large pretrained time-series models improve financial forecasting across global markets. Using over 10,000 assets spanning equities, bonds, and currencies, we compared ARIMA, XGBoost, LSTM, TimeGPT, and TimesFM under a rolling out-of-sample framework. We found that large pretrained models enhance cross-market stability and consistency, particularly in structurally stable markets such as Japan. However, gains in predictive accuracy were modest and highly conditional: XGBoost remained competitive in bonds and currencies, and no model dominated in efficient markets. Overall, the evidence suggests that pretrained models improve robustness rather than deliver universally superior forecasts, and that statistical improvements do not necessarily translate into economically meaningful advantages.
| [1] |
Akita R, Yoshihara A, Matsubara T, et al. (2016) Deep learning for stock prediction using numerical and textual information. In 2016 IEEE/ACIS 15th International Conference on Computer and Information Science, 1–6. Okayama, Japan. https://doi.org/10.1109/ICIS.2016.7550882 |
| [2] |
Ardekani AM, Bertz J, Bryce C, et al. (2024) FinSentGPT: A universal financial sentiment engine?. Int Rev Financ Anal 94: 103291. https://doi.org/10.1016/j.irfa.2024.103291 doi: 10.1016/j.irfa.2024.103291
|
| [3] |
Athanasopoulos G, Hyndman RJ, Kourentzes N, et al. (2017) Forecasting with temporal hierarchies. European Eur J Oper Res 262: 60–74. https://doi.org/10.1016/j.ejor.2017.02.046 doi: 10.1016/j.ejor.2017.02.046
|
| [4] |
Basak S, Kar S, Saha S, et al. (2019) Predicting the direction of stock market prices using tree-based classifiers. North Am J Econ Finance 47: 552–567. https://doi.org/10.1016/j.najef.2018.06.003 doi: 10.1016/j.najef.2018.06.003
|
| [5] |
Box GEP, Jenkins GM (1976) Time series analysis: Forecasting and control. Holden-Day. |
| [6] |
Box GEP, Jenkins GM, Reinsel GC (2015) Time series analysis: Forecasting and control. John Wiley & Sons. |
| [7] |
Cao L, Tay FEH (2001) Financial forecasting using support vector machines. Neural Comput Appl 10: 184–192. https://doi.org/10.1007/s005210170010 doi: 10.1007/s005210170010
|
| [8] |
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, 785–794. https://doi.org/10.1145/2939672.2939785 |
| [9] |
Chen Q, Kawashima H (2024) Stock price prediction using llm-based sentiment analysis. In 2024 IEEE International Conference on Big Data (BigData), 4846–4853. https://doi.org/10.1109/BigData62323.2024.10825946 |
| [10] |
Das A, Kong W, Leber A, et al. (2024) A decoder-only foundation model for time-series forecasting. ICML 235: 10148–10167. |
| [11] |
Di Persio L, Honchar O (2016) Artificial neural networks architectures for stock price prediction: Comparisons and applications. Int J Circuits Syst Signal Process 10: 403–413. |
| [12] |
Engle RF (2001) GARCH 101: The use of ARCH/GARCH models in applied econometrics. J Econ Perspect 15: 157–168. https://doi.org/10.1257/jep.15.4.157 doi: 10.1257/jep.15.4.157
|
| [13] |
Engle RF (1982) Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica 50: 987–1007. https://doi.org/10.2307/1912773 doi: 10.2307/1912773
|
| [14] |
Fischer T, Krauss C (2018) Deep learning with long short-term memory networks for financial market predictions. Eur J Oper Res 270: 654–669. https://doi.org/10.1016/j.ejor.2017.11.054 doi: 10.1016/j.ejor.2017.11.054
|
| [15] |
Garza A, Mergenthaler-Canseco M (2024) TimeGPT: Foundation models for time series forecasting. arXiv preprint. |
| [16] |
Graves A, Mohamed A, Hinton G (2013) Speech recognition with deep recurrent neural networks. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 6645–6649. https://doi.org/10.1109/ICASSP.2013.6638947 |
| [17] |
Greff K, Srivastava RK, Koutník J, et al. (2017) LSTM: A search space odyssey. IEEE Trans Neural Netw Learn Syst 28: 2222–2232. https://doi.org/10.1109/TNNLS.2016.2582924 doi: 10.1109/TNNLS.2016.2582924
|
| [18] |
Guresen E, Kayakutlu G, Daim TU (2011) Using artificial neural network models in stock market index prediction. Expert Syst Appl 38: 10389–10397. https://doi.org/10.1016/j.eswa.2011.02.068 doi: 10.1016/j.eswa.2011.02.068
|
| [19] |
Guo H, Kwok PY, Guo Y, et al. (2025) FinLSPM: Large stock predict model via numerical prior knowledge from LLM. Expert Syst Appl, 130294. https://doi.org/10.1016/j.eswa.2025.130294 doi: 10.1016/j.eswa.2025.130294
|
| [20] |
Hansen PR, Lunde A, Nason JM (2010) The model confidence set. Econometrica 78: 453–497. https://doi.org/10.3982/ECTA5771 ttps://doi.org/10.3982/ECTA5771 doi: 10.3982/ECTA5771
|
| [21] |
Hochreiter S, Schmidhuber J (1997) Long short-term memory. Neural Comput 9: 1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735 doi: 10.1162/neco.1997.9.8.1735
|
| [22] |
Hussain SA, Qiu-Shi C, Hussain SA, et al. (2026) Improving financial forecasting with a synergistic LLM-transformer architecture: A hybrid approach to stock price prediction. arXiv preprint arXiv: 2601.02878. https://doi.org/10.62762/TIS.2025.976754 |
| [23] |
Hyndman RJ, Athanasopoulos G (2018) Forecasting: Principles and practice. OTexts. |
| [24] |
Kamble RA (2017) Short and long term stock trend prediction using decision tree. In 2017 International Conference on Intelligent Computing and Control Systems, 1371–1375. https://doi.org/10.1109/ICCONS.2017.8250694 |
| [25] |
Kim HJ, Shin KS (2007) A hybrid approach based on neural networks and genetic algorithms for detecting temporal patterns in stock markets. Appl Soft Comput 7: 569–576. https://doi.org/10.1016/j.asoc.2006.03.002 doi: 10.1016/j.asoc.2006.03.002
|
| [26] |
Lim B, Zohren S (2021) Time series forecasting with deep learning: A survey. Philos Trans R Soc A 379: 20200209. https://doi.org/10.1098/rsta.2020.0209 doi: 10.1098/rsta.2020.0209
|
| [27] |
Li Y, Zhu Z, Kong D, et al. (2020) Temporal fusion transformers for interpretable multi-horizon time series forecasting. Int J Forecast 37: 1748–1764. https://doi.org/10.1016/j.ijforecast.2020.07.006 doi: 10.1016/j.ijforecast.2020.07.006
|
| [28] |
Malkiel BG, Fama EF (1970) Efficient capital markets: A review of theory and empirical work. J Finance 25: 383–417. https://doi.org/10.2307/2325486 doi: 10.2307/2325486
|
| [29] |
Makridakis S, Spiliotis E, Assimiakopoulos V (2022) Statistical, machine learning and deep learning forecasting methods: Comparisons and ways forward. J Oper Res Soc 73: 840–859. https://doi.org/10.1080/01605682.2022.2118629 doi: 10.1080/01605682.2022.2118629
|
| [30] |
Mutian O, Thomas JJ, Tianzhou Y, et al. (2025) LLM-guided semantic feature selection for interpretable financial market forecasting in low-resource financial markets. Franklin Open, 100359. https://doi.org/10.1016/j.fraope.2025.100359 doi: 10.1016/j.fraope.2025.100359
|
| [31] |
Nelson DM, Pereira AC, de Oliveira RA (2017) Stock market's price movement prediction with LSTM neural networks. In 2017 International Joint Conference on Neural Networks, 1419–1426. https://doi.org/10.1109/IJCNN.2017.7966030 |
| [32] |
Nobre J, Neves RF (2019) Combining principal component analysis, discrete wavelet transform and XGBoost to trade in the financial markets. Expert Syst Appl 125: 181–194. https://doi.org/10.1016/j.eswa.2019.01.076 doi: 10.1016/j.eswa.2019.01.076
|
| [33] |
Oreshkin BN, Carpov D, Chapados N, et al. (2020) N-BEATS: Neural basis expansion analysis for interpretable time series forecasting. International Conference on Learning Representations. |
| [34] |
Rather AM, Agarwal A, Sastry V (2015) Recurrent neural network and a hybrid model for prediction of stock returns. Expert Syst Appl 42: 3234–3241. https://doi.org/10.1016/j.eswa.2014.12.003 doi: 10.1016/j.eswa.2014.12.003
|
| [35] |
Rout AK, Dash P, Dash R, et al. (2015) Forecasting financial time series using a low complexity recurrent neural network and evolutionary learning approach. J King Saud Univ Comput Inf Sci 29: 536–552. https://doi.org/10.1016/j.jksuci.2015.06.003 doi: 10.1016/j.jksuci.2015.06.003
|
| [36] |
Selvin S, Vinayakumar R, Gopalakrishnan EA, et al. (2017) Stock price prediction using LSTM, RNN and CNN-sliding window model. In 2017 International Conference on Advances in Computing, Communications and Informatics (ICACCI), 1643–1647. https://doi.org/10.1109/ICACCI.2017.8126078 |
| [37] |
Shiller RJ (2003) From efficient markets theory to behavioral finance. J Econ Perspect 17: 83–104. https://doi.org/10.1257/089533003321164967 doi: 10.1257/089533003321164967
|
| [38] |
Sun Y, Gao P, Yan Y, et al. (2025) Stock price prediction with attention-based framework by integrating LLM-generated features. In International Conference on Neural Information Processing, 562–576.https://doi.org/10.1007/978-981-95-4445-5_38 |
| [39] |
Sureshkumar KK, Elango NM (2012) Performance analysis of stock price prediction using artificial neural network. GJCST 12: 19–25. |
| [40] |
Sutskever I, Vinyals O, Le QV (2014) Sequence to sequence learning with neural networks. NeurIPS 27: 3104–3112. https://doi.org/10.48550/arXiv.1409.3215 doi: 10.48550/arXiv.1409.3215
|
| [41] |
Tsay RS (2005) Analysis of financial time series (2nd ed.). Wiley. https://doi.org/10.1002/0471746193 |
| [42] |
Vaswani A, Shazeer N, Parmar N, et al. (2017) Attention is all you need. In Advances in Neural Information Processing Systems 30: 5998–6008. https://doi.org/10.48550/arXiv.1706.03762 |
| [43] |
Yoshihara A, Fujikawa K, Seki K, et al. (2014) Predicting stock market trends by recurrent deep neural networks. In Proc. Pacific Rim International Conference on Artificial Intelligence, 759–969. https://doi.org/10.1007/978-3-319-13560-1_60 |
| [44] |
Yu Y, Si X, Hu C, et al. (2019) A review of recurrent neural networks. Neural Comput Appl 31: 1233–1255. https://doi.org/10.1007/s00521-018-3628-5 doi: 10.1007/s00521-018-3628-5
|
| [45] |
Zhou L, Zhang Y, Yu J, et al. (2025) Llm-augmented linear transformer–cnn for enhanced stock price prediction. Mathematics 13: 487. https://doi.org/10.3390/math13030487 doi: 10.3390/math13030487
|
 QFE-10-04-028-S001.pdf |
 |
Figures(6) / Tables(4)
Linghan Huang, Yunhan Li, Xuan Feng. Do pretrained models improve financial time series forecasting?[J]. Quantitative Finance and Economics, 2026, 10(4): 728-752. doi: 10.3934/QFE.2026028
Cross-market robustness (RI) for basic materials sector across countries. Notes: Results are based on rolling out-of-sample forecasts. Higher RI indicates greater stability across assets and time.
Cross-market robustness (RI) for basic materials sector across broader regions. Notes: Results are based on rolling out-of-sample forecasts. Higher RI indicates greater stability across assets and time.
Cross-market robustness (RI) for financial services sector across countries. Notes: Results are based on rolling out-of-sample forecasts. Higher RI indicates greater stability across assets and time.
Cross-market robustness (RI) for financial services sector across broader regions. Notes: Results are based on rolling out-of-sample forecasts. Higher RI indicates greater stability across assets and time.
Sector-level RMSE comparison across models (Germany). Notes: Lower RMSE indicates higher predictive accuracy.
Sector-level RMSE comparison across models (United States). Notes: Lower RMSE indicates higher predictive accuracy.
DownLoad: