金融市場價格波動日益劇烈，準確預測股價的非線性特徵已成為財務工程與人工智慧領域的重要挑戰。隨著深度學習技術的快速發展，循環神經網路（RNN）相關模型已成為時間序列預測的有力工具，然而既有研究多聚焦於單一模型或特定產業的驗證，缺乏一個能夠系統性比較多種深度學習架構在各類股族群中預測準確度的綜合性研究。
故本研究以臺灣五大產業（半導體、光電、金融、汽車、航運）之上市公司為研究對象，選取各產業獲利前五大公司，研究區間涵蓋 2009 年至 2025 年，共計 4,356 個交易日。本研究建置了 RNN、GRU、LSTM 以及多重時間區間長短期記憶模型（MI-LSTM）共四種模型，透過一致的訓練參數設定，比較各模型在不同產業環境下的預測能力，並採用 MSE、MAE、RMSE 與 MAPE 四項誤差指標進行綜合評估，以檢視模型之適用性與跨產業穩健性，經實證分析得到以下結論：
四種模型間的預測準確度存在顯著差異，MI-LSTM 表現出更佳的趨勢追蹤與轉折預測能力。在資料區間中波動度最大的航運（43.14%）與光電（42.02%）中，MI-LSTM 模型憑藉其多時間尺度的平行輸入結構，仍能有效捕捉長短期波動特徵。
MI-LSTM 透過多重時間區間（Multi-Interval）結構，能同時捕捉單一價格變數在不同時間尺度下（短期衝擊、中期波段、長期趨勢）的非線性演進特徵，因而具備跨產業應用的穩健優勢。從各產業的模型排名觀察，MI-LSTM於五大產業中皆穩居第一，呈現極高的排序一致性，而RNN 模型普遍位居第二。MI-LSTM模型在股價高波動的半導體與航運產業之優勢尤為明顯；在股價相對平穩的金融保險中，MI-LSTM模型亦維持最低誤差，顯示模型具跨產業穩健性。

Accurately predicting non-linear stock prices amidst high market volatility is a critical challenge in financial engineering and AI. While RNN models excel at time-series forecasting, systematic comparisons across multiple architectures and industries remain scarce. This study analyzes 25 leading companies across five major Taiwanese sectors (Semiconductors, Optoelectronics, Finance, Automotive, and Shipping) from 2009 to 2025 (4,356 trading days). We evaluated RNN, GRU, LSTM, and Multi-Interval LSTM (MI-LSTM) using MSE, MAE, RMSE, and MAPE metrics to verify their cross-industry robustness.
Empirical results indicate:
1. Superior Performance: MI-LSTM significantly outperforms other models in trend tracking and pivot point prediction.
2. Volatility Impact: Prediction accuracy correlates negatively with industry volatility; however, MI-LSTM’s advantages are most pronounced in high-volatility sectors, suggesting that industrial characteristics systematically influence model performance.
3. Multi-Scale Structure and Volatility Adaptability: In Shipping and Optoelectronics—the two industries with the highest annualized volatility in the dataset (43.14% and 42.02%, respectively)—the MI-LSTM model, leveraging its multi-scale parallel input structure, successfully captures both short-term shocks and long-term evolutionary trends from a single price series.
4. Ranking Consistency: MI-LSTM ranked first across all sectors—notably in high-volatility Semiconductors and Shipping—while maintaining the lowest error in the stable Finance sector. RNN generally followed in second place.

一、英文文獻
1. Ang, A., and M. Piazzesi, “A No-Arbitrage Vector Autoregression of Term Structure Dynamics with Macroeconomic Variables,” Journal of Monetary Economics, Vol. 50, No. 4, pp. 745–787, 2003.
2. Bao, W., J. Yue, and Y. Rao, “A Deep Learning Framework for Financial Time Series Using Stacked Autoencoders and Long Short-Term Memory,” PLOS ONE, Vol. 12, No. 7, Article e0180944, 2017.
3. Bengio, Y., P. Simard, and P. Frasconi, “Learning Long-Term Dependencies with Gradient Descent Is Difficult,” IEEE Transactions on Neural Networks, Vol. 5, No. 2, pp. 157–166, 1994.
4. Black, F., “Noise,” The Journal of Finance, Vol. 41, No. 3, pp. 528–543, 1986.
5. Box, G. E. P., and G. M. Jenkins, Time Series Analysis: Forecasting and Control, revised ed., Holden-Day, San Francisco, CA, 575 pp., 1976.
6. Cho, K., B. van Merriënboer, C. Gulcehre, D. Bahdanau, F. Bougares, H. Schwenk, and Y. Bengio, “Learning Phrase Representations Using RNN Encoder–Decoder for Statistical Machine Translation,” Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing, pp. 1724–1734, 2014.
7. Chong, E., C. Han, and F. C. Park, “Deep Learning Networks for Stock Market Analysis and Prediction: Methodology, Data Representations, and Case Studies,” Expert Systems with Applications, Vol. 83, pp. 187–205, 2017.
8. Chung, J., C. Gulcehre, K. Cho, and Y. Bengio, “Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling,” arXiv Preprint, arXiv:1412.3555, 2014.
9. Easley, D., and M. O’Hara, “Information and the Cost of Capital,” The Journal of Finance, Vol. 59, No. 4, pp. 1553–1583, 2004.
10. Fama, E. F., “Efficient Capital Markets: A Review of Theory and Empirical Work,” The Journal of Finance, Vol. 25, No. 2, pp. 383–417, 1970.
11. Fama, E. F., and K. R. French, “A Five-Factor Asset Pricing Model,” Journal of Financial Economics, Vol. 116, No. 1, pp. 1–22, 2015.
12. Fischer, T., and C. Krauss, “Deep Learning with Long Short-Term Memory Networks for Financial Market Predictions,” European Journal of Operational Research, Vol. 270, No. 2, pp. 654–669, 2018.
13. Greff, K., R. K. Srivastava, J. Koutník, B. R. Steunebrink, and J. Schmidhuber, “LSTM: A Search Space Odyssey,” IEEE Transactions on Neural Networks and Learning Systems, Vol. 28, No. 10, pp. 2222–2232, 2017.
14. Gu, S., B. Kelly, and D. Xiu, “Empirical Asset Pricing via Machine Learning,” The Review of Financial Studies, Vol. 33, No. 5, pp. 2223–2273, 2020.
15. Hochreiter, S., and J. Schmidhuber, “Long Short-Term Memory,” Neural Computation, Vol. 9, No. 8, pp. 1735–1780, 1997.
16. Hong, H., W. Torous, and R. Valkanov, “Do Industries Lead the Stock Market?” Journal of Financial Economics, Vol. 83, No. 2, pp. 367–396, 2007.
17. Jozefowicz, R., W. Zaremba, and I. Sutskever, “An Empirical Exploration of Recurrent Network Architectures,” Proceedings of the 32nd International Conference on Machine Learning, Vol. 37, pp. 2342–2350, 2015.
18. Kim, E. B., J. H. Park, Y. S. Lee, and C. Lim, “Two-Dimensional Attention-Based Multi-Input LSTM for Time Series Prediction,” Communications for Statistical Applications and Methods, Vol. 28, No. 1, pp. 39–57, 2021.
19. Lim, B., S. Ö. Arık, N. Loeff, and T. Pfister, “Temporal Fusion Transformers for Interpretable Multi-Horizon Time Series Forecasting,” International Journal of Forecasting, Vol. 37, No. 4, pp. 1748–1764, 2021.
20. Lo, A. W., “The Adaptive Markets Hypothesis: Market Efficiency from an Evolutionary Perspective,” The Journal of Portfolio Management, Vol. 30, No. 5, pp. 15–29, 2004.
21. Menzly, L., and O. Ozbas, “Market Efficiency and Information Diffusion across Chains of Production,” The Journal of Finance, Vol. 65, No. 2, pp. 561–600, 2010.
22. Nelson, D. M. Q., A. C. M. Pereira, and R. A. de Oliveira, “Stock Market’s Price Movement Prediction with LSTM Neural Networks,” Proceedings of the 2017 International Joint Conference on Neural Networks, pp. 1419–1426, 2017.
23. Pascanu, R., T. Mikolov, and Y. Bengio, “On the Difficulty of Training Recurrent Neural Networks,” Proceedings of the 30th International Conference on Machine Learning, Vol. 28, pp. 1310–1318, 2013.
24. Qiu, J., B. Wang, and C. Zhou, “Forecasting Stock Prices with Long-Short Term Memory Neural Network Based on Attention Mechanism,” PLOS ONE, Vol. 15, No. 1, Article e0227222, 2020.
25. Sezer, O. B., M. U. Gudelek, and A. M. Ozbayoglu, “Financial Time Series Forecasting with Deep Learning: A Systematic Literature Review, 2005–2019,” Applied Soft Computing, Vol. 90, Article 106181, 2020.
26. Shi, X., Z. Chen, H. Wang, D. Y. Yeung, W. K. Wong, and W. C. Woo, “Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting,” Advances in Neural Information Processing Systems, Vol. 28, pp. 802–810, 2015.
27. Siami-Namini, S., N. Tavakoli, and A. S. Siami Namin, “A Comparison of ARIMA and LSTM in Forecasting Time Series,” Proceedings of the 17th IEEE International Conference on Machine Learning and Applications, pp. 1394–1401, 2018.
28. Sirignano, J., and R. Cont, “Universal Features of Price Formation in Financial Markets: Perspectives from Deep Learning,” Quantitative Finance, Vol. 19, No. 9, pp. 1449–1459, 2019.
29. Stopford, M., Maritime Economics, 3rd ed., Routledge, London, 840 pp., 2009.
二、中文文獻
1. 粘凱婷、金成隆、周濟群與汪戊安，〈法人說明會資訊在供應鏈中之垂直資訊移轉效果：以台灣之半導體產業供應鏈為例〉，《臺大管理論叢》，第26卷，第3期，頁1–34，2016年。