人們對於股價預測深感興趣，然而身處在大數據世代，雖然有多元便利的資訊來源，但卻對多元的變因，使得預測股價仍是一件燙手山芋的問題。在此研究中，首先我們先收集並有效預處理歷史交易資訊、技術指標、基本面財務報表。再者，我們透過多工的雙重注意力機制，針對所有輸入特徵進行雙重注意力機制，以利我們能有效抽取輸入特徵及時間相依之間的關聯性，以預測股價實際漲跌數值。最後，結果顯示模型之預測效果遠比過去研究平均提高 94% 的效果，並伴隨著最低的誤差。藉由我們提出的模型，不僅透過適當捕捉長期時間的相依性去有效擷取相關特徵，也可以善用技術指標跟基本面會計知識，以預測最接近真實的股票價格。除了客觀實驗結果，我們也透過問卷調查，受試者包含多位業界專業投資部門主管，希望藉由人類投資行為的主觀評估調查，來協助我們比較人類與AI之間的邏輯思維差異，供我們進行深入探討。

People always are fascinated about stock price forecasting. However, in the era of the big data, even if we have many kinds of information resource, it is still a difficult issue for stock price prediction due to multiple variables. About our research, at first, we collect and preprocess efficiently historical trading prices, technical indicators, and fundamental accounting information. Then, through multiple dual stage attention mechanism, our proposed Technical Fundamental_Dual Attention (TF_DA) model could efficiently extract the co-relationship between input features and temporal dependencies in order to predict stock prices. Finally, the result proves that our model could effectively improve 94% of the performance on average with the least error than past research. Our model could extract both of the relevant input features and temporal dependencies in the long term. Besides, it could also take advantage of the information about technical indicators and fundamental accounting items to predict the values closest to stock prices. In addition to the objective evaluation of our model, we also have the subject evaluation for human investment behavior. The subjects indeed include professional investment managers. Through the questionnaire, it could help us to discuss the logical thinking difference between human and artificial intelligence.

摘要
Abstract
Acknowledgement
List of Tables
List of Figures
Introduction--------------------------------1
Related Work--------------------------------5
Methodology---------------------------------25
Experiments and Results---------------------51
Conclusion and Future Work------------------70
References----------------------------------72
A Technical Indicators----------------------79
B Fundamental Accounting Items--------------89
C Total Alpha Attention of Top 20 Weights---95

[1] Eugene F Fama. Efficient capital markets: A review of theory and empirical work.The Journal of Finance, 25(2):383–417, 1970. ISSN 00221082, 15406261. URL: https://www.jstor.org/stable/2325486.
[2] Andrew W Lo. The adaptive markets hypothesis.The Journal of Portfolio Manage-ment, 30(5):15–29, 2004. doi: 10.3905/jpm.2004.442611.
[3] C. R. Cavalcante, C. R. Brasileiro, L. F. V. Souza, P. J. Nobrega, and A. L. Oliveira.Computational intelligence and financial markets: A survey and future directions.Expert Systems with Applications, 55:194–211, 2016.
[4] G. S. Atsalakis and K. P. Valavanis. Surveying stock market forecasting techniques -part ii: Machine learning methods.Expert Systems with Applications, 36:5932–5941,2009.
[5] B. Krollner, B. Vanstone, and G. Finnie. Financial time series forecasting with ma-chine learning techniques: A survey.European symposium on artificial neural net-works: Computational and machine learning, 2010.
[6] B. Vanstone and C. Tan. A survey of the application of soft computing to investment72 and financial trading.Proceedings of the Australian and New Zealand IntelligentInformation Systems Conference, pages 211–216, 2003.
[7] William Brock, Josef Lakonishok, and Blake Lebaron. Simple technical trading rulesand the stochastic properties of stock returns.The Journal of Financial, 47(5):1731–1764, 1992. ISSN 1540-6261. doi: 10.1111/j.1540-6261.1992.tb04681.x.
[8] Rodolfo Torbio Farias Nazrio, Jssica Lima e Silva, Vinicius Amorim Sobreiro, andHerbert Kimura. A literature review of technical analysis on stock markets.TheQuarterly Review of Economics and Finance, 2017. ISSN 1062-9769. doi: 10.11016/j.qref.2017.01.014.
[9] Baruch Lev and S Ramu Thiagarajan. Fundamental information analysis.Journal ofAccounting research, pages 190–215, 1993.
[10] Patricia M Dechow, Amy P Hutton, Lisa Meulbroek, and Richard G Sloan. Short-sellers, fundamental analysis, and stock returns.Journal of Financial Economics,61(1):77–106, 2001. doi: 10.1016/S0304-405X(01)00056-3. URL: http://www.sciencedirect.com/science/article/pii/S0304405X01000563.
[11] Schumaker, R. P., Zhang, Y. , Huang, C.N., and Chen, H. . Evaluating sentiment infinancial news articles.Decision Support Systems, 53(3):458–464, 2012.
[12] Nassirtoussi, A. K., Aghabozorgi, S., Wah, T. Y., and Ngo, D. C. L. Text miningof news-headlines for forex market prediction: A multi-layer dimension reductionalgorithm with semantics and sentiment.Expert Systems with Applications, 42(1):306–324, 2015.
[13] P. Whittle. Hypothesis testing in time series analysis.PhD thesis, 1951.
[14] Dimitros Asteriou and Stephen G Hall. ARIMA models and the box-jenkins method-ology.Applied Econometrics, 2(2):265–286, 2011.
[15] Eugen Diaconescu. The use of NARX neural networks to predict chaotic time series.WSEA Transactions on Computer Research, 3(3), 2008.
[16] Linjun Yan, Ahmed Elgamal, and Garrison W. Cottrell. Substructure vibration NARXneural network approach for statistical damage inference.Journal of EngineeringMechanics, 139:737–747, 2013.
[17] S. Chen, X. X. Wang, and C. J. Harris. NARX-based nonlinear system identificationusing orthogonal least squares basis hunting.IEEE Transactions on Control SystemsTechnology, 16(1):78–84, 2008.
[18] Abdelhamid Bouchachia and Saliha Bouchachia. Ensemble learning for time seriesprediction.In Proceedings of the 1st International Workshop on Nonlinear Dynamicsand Synchronization, 2008.
[19] R. Frigola and C. E. Rasmussen. Integrated pre-processing for bayesian nonlinearsystem identification with gaussian processes.IEEE Conference on Decision andControl, pages 552–560, 2014.
[20] Y. LeCun, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard, andL. D. Jackel. Backpropagation applied to handwritten zip code recognition.NeuralCompute, 1(4):541–551, 1989. ISSN 0899-7667. doi: 10.1162/neco.1989.1.4.541. URL: http://dx.doi.org/10.1162/neco.1989.1.4.541.
[21] Omer Berat Sezer and Ahmet Murat Ozbayoglu. Algorithmic financial trading withdeep convolutional neural networks: Time series to image conversion approach.Ap-plied Soft Computing, 70:525–538, 2018.
[22] Shuanglong Liu, Chao Zhang, and Jinwen Ma. CNN-LSTM neural network modelfor quantitative strategy analysis in stock markets.Neural Information Processing,pages 198–206, 2017.
[23] Jerzy Korczak and Marcin Hernes. Deep learning for financial time series forecast-ing in a trader system.Proceedings of the 2017 Federated Conference on ComputerScience and Information Systems, 2017.
[24] Narek Abroyan. Neural networks for financial market risk classification.Frontiers inSignal Processing, 1(2), 2017.
[25] David E Rumelhart, Geoffrey E Hinton, and Ronald J Williams. Learning represen-tations by backpropagating errors.Nature, 323(9):533–536, 1986.
[26] Paul J Werbos. Backpropagation through time: what it does and how to do it.Pro-ceedings of the IEEE, 78(10):1550–1560, 1990.
[27] Jeffrey L Elman. Distributed representations, simple recurrent networks, and gram-matical structure.Machine learning, 7(2–3):195–225, 1991.
[28] Yang Gao and Meng Joo Er. NARMAX time series model prediction: feedforwardand recurrent fuzzy neural network approaches.Fuzzy Sets and Systems, 150(2):331–350, 2005.
[29] Bruno Spilak. Deep neural networks for cryptocurrencies price prediction.Masterthesis, HumboldtUniversitat zu Berlin, Wirtschaftswissenschaftliche Fakultat, 2018.
[30] E.W. Saad, D.V. Prokhorov, and D.C. Wunsch. Comparative study of stock trendprediction using time delay, recurrent and probabilistic neural networks.IEEE Trans-actions on Neural Networks, 9(6):1456–1470, 1998.
[31] P. Tino, C. Schittenkopf, and G. Dorffner. Financial volatility trading using recurrentneural networks.IEEE Transactions on Neural Networks, 12(4):865–874, 2001.
[32] Yoshua Bengio, Patrice Simard, and Paolo Frasconi. Learning long-term dependen-cies with gradient descent is difficult.IEEE Transactions on Neural Networks, 5(2):157–166, 1994.
[33] Sepp Hochreiter and Jurgen Schmidhuber. Long short-term memory.Neural Compu-tation, 9(8):1735–1780, 1997.
[34] Kyunghyun Cho, Bart Van Merrienboer, Caglar Gulcehre, Dzmitry Bahdanau, FethiBougares, Holger Schwenk, and Yoshua Bengio. Learning phrase representationsusing RNN encoder-decoder for statistical machine translation. 2014. URL: https://arxiv.org/abs/1406.1078.
[35] Ilya Sutskever, Oriol Vinyals, and Quoc V Le. Sequence to sequence learning withneural networks.NIPS, pages 3104–3112, 2014.
[36] Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, AidanN. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. 2017. URL: https://arxiv.org/abs/1706.03762.
[37] Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. Neural machine translationby jointly learning to align and translate. 2014. URL: https://arxiv.org/abs/1409.0473.
[38] Guokun Lai, Wei-Cheng Chang, Yiming Yang, and Hanxiao Liu. Modeling long- andshort-term temporal patterns with deep neural networks.SIGIR, 2018.
[39] Shun-Yao Shih, Fan-Keng Sun, and Hung yi Lee. Temporal pattern attention formultivariate time series forecasting.Journal Track of the European Conference onMachine Learning and Principles and Practice of Knowledge Discovery in Databases(ECMLPKDD), 2019. URL: https://arxiv.org/abs/1809.04206.
[40] E. Beyaz, F. Tekiner, X. Zeng, and J. Keane. Comparing technical and fundamen-tal indicators in stock price forecasting.2018 IEEE 20th International Confer-ence on High Performance Computing and Communications; IEEE 16th Interna-tional Conference on Smart City; IEEE 4th International Conference on Data Sci-ence and Systems (HPCC/SmartCity/DSS), pages 1607–1613, 2018. doi: 10.1109/HPCC/SmartCity/DSS.2018.00262. URL: https://ieeexplore.ieee.org/document/8623000.
[41] XingYu Fu, JinHong Du, YiFeng Guo, MingWen Liu, Tao Dong, and XiuWen Duan.A machine learning framework for stock selection.National Sun Yat-sen University,2018. URL: https://arxiv.org/abs/1806.01743.77.
[42] Yao Qin, Dongjin Song, Haifeng Chen, Wei Cheng, Guofei Jiang, and Garrison Cot-trell. A dual-stage attention-based recurrent neural network for time series predic-tion.International Joint Conference on Artificial Intelligence (IJCAI), 2017. URL: https://arxiv.org/abs/1704.02971.
[43] Linyi Yang, Zheng Zhang, Su Xiong, Lirui Wei, James Ng, Lina Xu, and RuihaiDong. Explainable text-driven neural network for stock prediction.2018 5th IEEEInternational Conference on Cloud Computing and Intelligence Systems, 2018. URL: https://arxiv.org/abs/1902.04994.
[44] Siteng Huang, Donglin Wang, Xuehan Wu, and Ao Tang. Dsanet: Dual self-attentionnetwork for multivariate time series forecasting.The 28th ACM International Con-ference on Information and Knowledge Management (CIKM 2019), 2019. URL: https://dl.acm.org/doi/10.1145/3357384.3358132.
[5] B. Krollner, B. Vanstone, and G. Finnie. Financial time series forecasting with ma-chine learning techniques: A survey.European symposium on artificial neural net-works: Computational and machine learning, 2010.
[6] B. Vanstone and C. Tan. A survey of the application of soft computing to investment72 and financial trading.Proceedings of the Australian and New Zealand IntelligentInformation Systems Conference, pages 211–216, 2003.
[7] William Brock, Josef Lakonishok, and Blake Lebaron. Simple technical trading rulesand the stochastic properties of stock returns.The Journal of Financial, 47(5):1731–1764, 1992. ISSN 1540-6261. doi: 10.1111/j.1540-6261.1992.tb04681.x.
[8] Rodolfo Torbio Farias Nazrio, Jssica Lima e Silva, Vinicius Amorim Sobreiro, andHerbert Kimura. A literature review of technical analysis on stock markets.TheQuarterly Review of Economics and Finance, 2017. ISSN 1062-9769. doi: 10.11016/j.qref.2017.01.014.
[9] Baruch Lev and S Ramu Thiagarajan. Fundamental information analysis.Journal ofAccounting research, pages 190–215, 1993.
[10] Patricia M Dechow, Amy P Hutton, Lisa Meulbroek, and Richard G Sloan. Short-sellers, fundamental analysis, and stock returns.Journal of Financial Economics,61(1):77–106, 2001. doi: 10.1016/S0304-405X(01)00056-3. URL: http://www.sciencedirect.com/science/article/pii/S0304405X01000563.
[11] Schumaker, R. P., Zhang, Y. , Huang, C.N., and Chen, H. . Evaluating sentiment infinancial news articles.Decision Support Systems, 53(3):458–464, 2012.
[12] Nassirtoussi, A. K., Aghabozorgi, S., Wah, T. Y., and Ngo, D. C. L. Text miningof news-headlines for forex market prediction: A multi-layer dimension reductionalgorithm with semantics and sentiment.Expert Systems with Applications, 42(1):306–324, 2015.
[13] P. Whittle. Hypothesis testing in time series analysis.PhD thesis, 1951.
[14] Dimitros Asteriou and Stephen G Hall. ARIMA models and the box-jenkins method-ology.Applied Econometrics, 2(2):265–286, 2011.
[15] Eugen Diaconescu. The use of NARX neural networks to predict chaotic time series.WSEA Transactions on Computer Research, 3(3), 2008.[16] Linjun Yan, Ahmed Elgamal, and Garrison W. Cottrell. Substructure vibration NARXneural network approach for statistical damage inference.Journal of EngineeringMechanics, 139:737–747, 2013.
[17] S. Chen, X. X. Wang, and C. J. Harris. NARX-based nonlinear system identificationusing orthogonal least squares basis hunting.IEEE Transactions on Control SystemsTechnology, 16(1):78–84, 2008.
[18] Abdelhamid Bouchachia and Saliha Bouchachia. Ensemble learning for time seriesprediction.In Proceedings of the 1st International Workshop on Nonlinear Dynamicsand Synchronization, 2008.
[19] R. Frigola and C. E. Rasmussen. Integrated pre-processing for bayesian nonlinearsystem identification with gaussian processes.IEEE Conference on Decision andControl, pages 552–560, 2014.
[20] Y. LeCun, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard, andL. D. Jackel. Backpropagation applied to handwritten zip code recognition.NeuralCompute, 1(4):541–551, 1989. ISSN 0899-7667. doi: 10.1162/neco.1989.1.4.541.URLhttp://dx.doi.org/10.1162/neco.1989.1.4.541.
[21] Omer Berat Sezer and Ahmet Murat Ozbayoglu. Algorithmic financial trading withdeep convolutional neural networks: Time series to image conversion approach.Ap-plied Soft Computing, 70:525–538, 2018
[22] Shuanglong Liu, Chao Zhang, and Jinwen Ma. CNN-LSTM neural network modelfor quantitative strategy analysis in stock markets.Neural Information Processing,pages 198–206, 2017.
[23] Jerzy Korczak and Marcin Hernes. Deep learning for financial time series forecast-ing in a trader system.Proceedings of the 2017 Federated Conference on ComputerScience and Information Systems, 2017.
[24] Narek Abroyan. Neural networks for financial market risk classification.Frontiers inSignal Processing, 1(2), 2017.
[25] David E Rumelhart, Geoffrey E Hinton, and Ronald J Williams. Learning represen-tations by backpropagating errors.Nature, 323(9):533–536, 1986.
[26] Paul J Werbos. Backpropagation through time: what it does and how to do it.Pro-ceedings of the IEEE, 78(10):1550–1560, 1990.
[27] Jeffrey L Elman. Distributed representations, simple recurrent networks, and gram-matical structure.Machine learning, 7(2–3):195–225, 1991.
[28] Yang Gao and Meng Joo Er. NARMAX time series model prediction: feedforwardand recurrent fuzzy neural network approaches.Fuzzy Sets and Systems, 150(2):331–350, 2005.
[29] Bruno Spilak. Deep neural networks for cryptocurrencies price prediction.Masterthesis, HumboldtUniversitat zu Berlin, Wirtschaftswissenschaftliche Fakultat, 2018.
[30] E.W. Saad, D.V. Prokhorov, and D.C. Wunsch. Comparative study of stock trendprediction using time delay, recurrent and probabilistic neural networks.IEEE Trans-actions on Neural Networks, 9(6):1456–1470, 1998.
[31] P. Tino, C. Schittenkopf, and G. Dorffner. Financial volatility trading using recurrentneural networks.IEEE Transactions on Neural Networks, 12(4):865–874, 2001.
[32] Yoshua Bengio, Patrice Simard, and Paolo Frasconi. Learning long-term dependen-cies with gradient descent is difficult.IEEE Transactions on Neural Networks, 5(2):157–166, 1994.
[33] Sepp Hochreiter and Jurgen Schmidhuber. Long short-term memory.Neural Compu-tation, 9(8):1735–1780, 1997.
[34] Kyunghyun Cho, Bart Van Merrienboer, Caglar Gulcehre, Dzmitry Bahdanau, FethiBougares, Holger Schwenk, and Yoshua Bengio. Learning phrase representationsusing RNN encoder-decoder for statistical machine translation. 2014. URL: https://arxiv.org/abs/1406.1078.
[35] Ilya Sutskever, Oriol Vinyals, and Quoc V Le. Sequence to sequence learning withneural networks.NIPS, pages 3104–3112, 2014.
[36] Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, AidanN. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. 2017. URL: https://arxiv.org/abs/1706.03762.
[37] Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. Neural machine translationby jointly learning to align and translate. 2014. URL: https://arxiv.org/abs/1409.0473.
[38] Guokun Lai, Wei-Cheng Chang, Yiming Yang, and Hanxiao Liu. Modeling long- andshort-term temporal patterns with deep neural networks.SIGIR, 2018.
[39] Shun-Yao Shih, Fan-Keng Sun, and Hung yi Lee. Temporal pattern attention formultivariate time series forecasting.Journal Track of the European Conference onMachine Learning and Principles and Practice of Knowledge Discovery in Databases(ECMLPKDD), 2019. URL: https://arxiv.org/abs/1809.04206.
[40] E. Beyaz, F. Tekiner, X. Zeng, and J. Keane. Comparing technical and fundamen-tal indicators in stock price forecasting.2018 IEEE 20th International Confer-ence on High Performance Computing and Communications; IEEE 16th Interna-tional Conference on Smart City; IEEE 4th International Conference on Data Sci-ence and Systems (HPCC/SmartCity/DSS), pages 1607–1613, 2018. doi: 10.1109/HPCC/SmartCity/DSS.2018.00262. URLhttps://ieeexplore.ieee.org/document/8623000.
[41] XingYu Fu, JinHong Du, YiFeng Guo, MingWen Liu, Tao Dong, and XiuWen Duan.A machine learning framework for stock selection.National Sun Yat-sen University,2018. URL: https://arxiv.org/abs/1806.01743.77
[42] Yao Qin, Dongjin Song, Haifeng Chen, Wei Cheng, Guofei Jiang, and Garrison Cot-trell. A dual-stage attention-based recurrent neural network for time series predic-tion.International Joint Conference on Artificial Intelligence (IJCAI), 2017. URL: https://arxiv.org/abs/1704.02971.
[43] Linyi Yang, Zheng Zhang, Su Xiong, Lirui Wei, James Ng, Lina Xu, and RuihaiDong. Explainable text-driven neural network for stock prediction.2018 5th IEEEInternational Conference on Cloud Computing and Intelligence Systems, 2018. URL: https://arxiv.org/abs/1902.04994.
[44] Siteng Huang, Donglin Wang, Xuehan Wu, and Ao Tang. Dsanet: Dual self-attentionnetwork for multivariate time series forecasting.The 28th ACM International Con-ference on Information and Knowledge Management (CIKM 2019), 2019. URL: https://dl.acm.org/doi/10.1145/3357384.3358132.