本研究以臺灣加權股價指數(TAIEX)為研究對象，探討融合新聞情緒與技術指標之神經網路模型於大盤漲跌趨勢預測的效果。研究主要分析ARIMA-GARCH模型捕捉之波動特徵，以及透過RoBERTa預訓練語言模型量化之財經新聞情緒，連同多項技術指標對大盤走勢預測準確率的影響。透過Investing.com與台灣新聞智慧網收集相關數據，並使用門控循環單元(GRU)結合滾動視窗法進行樣本外之實證檢驗。
研究結果顯示，混合模型的整體方向預測準確率達58.81%，在統計上顯著優於隨機猜測水準，顯現其在捕捉大盤非線性變動趨勢方面具正向影響。而在平穩的市場環境中，加入情緒與技術指標的混合模型並未與預測績效的提升呈顯著關聯。然而在市場恐慌期間，混合模型的預測準確率反而顯著提升至61.09%，顯示在極端不確定性下，投資人極易受情緒感染與從眾效應驅動，此時新聞情緒往往比傳統指標具備更佳的預測能力。
本研究補足過往文獻在總體市場指數預測與資訊方面之缺口，提供市場參與者在面臨極端市場變化時動態調整交易策略的參考。實證結果可作為投資者建構混合型量化模型以強化下檔風險保護之依據，亦建議機構與一般交易人將散戶情緒納入風險監測指標，用以評估並迴避恐慌性拋售之實質風險。

This study investigates a hybrid neural network integrating news sentiment and technical indicators to forecast the Taiwan Capitalization Weighted Stock Index (TAIEX). The model combines ARIMA-GARCH to extract volatility clustering, RoBERTa to quantify financial news sentiment, and technical indicators as features for a Gated Recurrent Unit network using a rolling-window approach for out-of-sample testing.
Empirical results show the hybrid model achieves an overall directional accuracy of 58.81%, significantly outperforming random guessing in capturing non-linear trends. While adding sentiment and technical features does not significantly improve performance in stable markets, the model's accuracy surges to 61.09% during market panics. This indicates that under extreme uncertainty, investor herding makes news sentiment a stronger predictor than traditional price-volume data.
This research addresses literature gaps in macro-index forecasting and information state-dependency. The findings provide actionable insights for investors to dynamically adjust trading strategies, build hybrid models for downside protection, and incorporate retail sentiment into risk monitoring to mitigate risks during panic sell-offs.

1. Aljifri, R. (2024). Herding during different types of crises: The COVID-19 health crisis and Russia–Ukraine political crisis. Heliyon, 10(10), e31382.
2. Antweiler, W., & Frank, M. Z. (2004). Is All That Talk Just Noise? The Information Content of Internet Stock Message Boards. The Journal of Finance, 59(3).
3. Bartov, E., Faurel, L., & Mohanram, P. S. (2018). Can Twitter help predict firm-level earnings and stock returns? The Accounting Review, 93(3), 25–57.
4. Broadstock, D. C., & Zhang, D. (2019). Social-media sentiment and Chinese stock markets: Coherency and directionality of information transmission. Journal of International Financial Markets, Institutions and Money, 58, 441–456.
5. Chiang, T. C., & Zheng, D. (2010). Herding in global stock markets. Journal of Banking & Finance, 34(8), 1911–1921.
6. Cho, K., van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning phrase representations using RNN encoder–decoder for statistical machine translation. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) (pp. 1724–1734).
7. Choi, S., & Kim, J. (2020). Herding behavior and investor sentiment: Evidence from the Korean stock market. Finance Research Letters, 32, 101166.
8. Cui, Y., Che, W., Liu, T., Qin, B., Wang, S., & Hu, G. (2020). Revisiting pre-trained models for Chinese natural language processing. In Findings of the Association for Computational Linguistics: EMNLP 2020 (pp. 657–668). Association for Computational Linguistics.
9. Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) (pp. 4171–4186).
10. Frijns, B., & Huynh, T. D. (2018). Herding in analyst recommendations: The role of media coverage and sentiment. Journal of Banking & Finance, 91, 1–18.
11. Gu, C., & Kurov, A. (2020). Informational role of social media: Evidence from Twitter sentiment. Journal of Banking & Finance, 121, 105969.
12. Huang, B.-N., & Yang, C.-W. (2001). An empirical investigation of trading volume and return volatility of the Taiwan Stock Market. Global Finance Journal, 12(1), 55–77.
13. Huang, T.-C., & Wang, K.-Y. (2017). Investors’ fear and herding behavior: Evidence from the Taiwan stock market. Emerging Markets Finance and Trade, 53(10), 2259–2278.
14. Huang, X., You, P., Gao, X., & Cheng, D. (2023). Stock price prediction based on ARIMA-GARCH and LSTM. In Proceedings of the 2023 International Conference on Business and Information Systems (ICBIS) (pp. 438–448).
15. Kulshreshtha, S., & Vijayalakshmi, A. (2020). An ARIMA-LSTM hybrid model for stock market prediction using live data. Journal of Engineering Science and Technology Review, 13(4), 117–123.
16. Kumar, B., Sunil, & Yadav, N. (2023). A hybrid approach for high-frequency time series forecasting: 𝛽SARMA-LSTM model. Journal of Forecasting, 42(8), 2089–2109.
17. Lahboub, K., & Benali, M. (2024). Assessing the predictive power of transformers, ARIMA, and LSTM in forecasting stock prices of Moroccan credit companies. Journal of Risk and Financial Management, 17(7), 293.
18. Liang, L. (2024). ARIMA with attention-based CNN-LSTM and XGBoost hybrid model for stock prediction in the US stock market. SHS Web of Conferences, 196, 02001.
19. Lyu, H. (2025). Hybrid ARIMA-LSTM model for stock market prediction: A time series and deep learning integration approach. Informatica, 49(22). https://doi.org/10.31449/inf.v49i22.8510
20. MacKinnon, J. G. (1996). Numerical distribution functions for unit root and cointegration tests. Journal of Applied Econometrics, 11(6), 601–618.
21. Munkh-Ulzii, B. J., McAleer, M., Moslehpour, M., & Wong, W.-K. (2018). Confucius and herding behaviour in the stock markets in China and Taiwan. Sustainability, 10(12), 4413.
22. Nelson, D. M. Q., Pereira, A. C. M., & de Oliveira, R. A. (2017). Stock market’s price movement prediction with LSTM neural networks. In 2017 International Joint Conference on Neural Networks (IJCNN) (pp. 1419–1426). IEEE.
23. Pabuccu, H., & Barbu, A. (2023). Feature selection with annealing for forecasting financial time series. arXiv. https://arxiv.org/abs/2303.02223
24. Sezer, O. B., Gudelek, M. U., & Ozbayoglu, A. M. (2020). Financial time series forecasting with deep learning: A systematic literature review: 2005–2019. Applied Soft Computing, 90, 106181.
25. Shi, Y., Wang, Y., Qu, Y. & Chen, Z. (2024), Integrated gcn-lstm stock prices movement prediction based on knowledge-incorporated graphs construction, International Journal of Machine Learning and Cybernetics 15(1), 161–176.
26. Shiller, R. J. (1981). Do stock prices move too much to be justified by subsequent changes in dividends? American Economic Review, 71(3), 421–436.
27. Si, Y., Nadarajah, S., Zhang, Z., & Xu, C. (2024). Modeling opening price spread of Shanghai Composite Index based on ARIMA-GRU/LSTM hybrid model. PLOS ONE, 19(3), e0299164.
28. Su, J., Lau, R. Y. K., Du, Y., Yu, J., & Zhang, H. (2025). A novel hybrid framework for stock price prediction integrating adaptive signal decomposition and multi-scale feature extraction. Applied Sciences, 15(23), 12450.
29. Sunki, A., SatyaKumar, C., Surya Narayana, G., Koppera, V., & Hakeem, M. (2024). Time series forecasting of stock market using ARIMA, LSTM and FB prophet. MATEC Web of Conferences, 392, Article 01163.
30. Tan, Q., Rasheed, M. H., & Rasheed, M. S. (2024). Post-COVID-19 technology adoption and noise trading: Elucidation of investors’ sentiments across cultures. China Accounting and Finance Review, 26(4), 431–458.
31. Tetlock, P. C. (2007). Giving content to investor sentiment: The role of media in the stock market. The Journal of Finance, 62(3), 1139–1168.
32. Yang, X. (2025). Prediction of Tencent share price based on ARIMA, SVR and LSTM. In Proceedings of the 1st International Conference on E-commerce and Artificial Intelligence (ECAI 2024) (pp. 229–233).
33. Yin, H. (2023). Enhancing directional accuracy in stock closing price value prediction using a direction-integrated MSE loss function. In Proceedings of the 1st International Conference on Data Analysis and Machine Learning (DAML 2023) (pp. 119–126).
34. Zhang, C., Li, Y., Chen, X., Jin, Y., Tang, P., & Li, J. (2020). DoubleEnsemble: A new ensemble method based on sample reweighting and feature selection for financial data analysis. arXiv. https://arxiv.org/abs/2010.01265