通過預測未來股票價格或是指數，例如開盤價或是收盤價，我們可以提前決定作多或是作空。除了股票指數的數值之外，對收盤價和開盤價之差的正值或是負值的預測是獲得利潤的重要訊息。本文提出了一種基於遞迴神經網路的方法來預測開盤價、收盤價以其兩者數值的差。與基於機器學習的現有方法相比，我們的方法鄭家專注於預處理，例如正規化的一階差分以及分析股票數據特性如過零率;一種代表了數據的符號在一個時間間隔內的變化率。我們提出了一種基於過零率估計的決策方法，以提高預測開盤價與收盤價之差的能力。我們將我們的方法應用於標準普爾500指數和道瓊工業指數。結果表明，我們的方法可以比以前的研究取得更好的結果。

By predicting the future stock price or index, such as the opening price and the closing price, we can place the long or short positions in advance. In addition to the stock index value, prediction on the positive or negative value of the difference between the closing price and the opening price is an important information for earning the profit. This paper presents a Recurrent Neural Network (RNN) based approach to forecast the opening price, the closing price and their difference. Compared to prior methods based on machine learning, our method puts greater focus on the pre-processing, such as normalized first order difference method, and the characteristics of the stock data, such as the zero-crossing rate (ZCR), which represents the ratio of data sign changes within a time interval. We propose a decision-making method based on an estimate of the ZCR to enhance the ability to predict the difference between the opening price and closing price. We apply our method to the Standard & Poor's 500 (S&P500) and the DowJones stock index. The results indicate that
our method can achieve better outcomes than prior works.

第一章 簡介 page 1
第二章 遞迴神經網路以及股市走勢的回顧 page 5
2.1 遞迴神經網路 page 5
2.2 長短期記憶 page 8
2.3 網路最佳化以及產生訓練集 page 10
2.4 股市中的價差 page 12
第三章 建立模型架構以及預測股市開盤、收盤以其價差 page 13
3.1 透過正規化之一階導數進行資料預處理以及預測 page 15
3.2 模型參數設定以及評估方法 page 17
3.3 針對股市價格的預測模擬評估結果 page 20
第四章 建立模型架構以及針對價差預測漲跌 page 29
4.1 利用股票趨勢以及過零率的決策方法 page 31
4.2 利用滑動窗口來估計過零率 page 34
4.3 針對漲跌的預測模擬評估 page 36
第五章 結論 page 42

[1] I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. USA: MIT Press: Cambridge, MA, 2016.
[2] S. Hochreiter and J. Schmidhuber, "Long short-term memory," Neural Computation, vol. 9, no. 8, pp. 1735-1780, November 1997.
[3] I. Aldridge, High-Frequency Trading: A Practical Guide to Algorithmic Strategies and Trading Systems. Wiley, 2013.
[4] S. M. Chen and C. D. Chen, "Taiex forecasting based on fuzzy time series and fuzzy variation groups," IEEE Transactions on Fuzzy Systems, vol. 19, no. 1, pp. 1-12, Feb 2011.
[5] E. Hadavandi ; H. Shavandi ; A. Ghanbari, "A genetic fuzzy expert system for stock price forecasting," in Proc. IEEE International Conference on Fuzzy Systems and Knowledge Discovery (FSKD), vol. 1, Cambridge, UK, August 2010, pp. 41-44.
[6] A. A. Ariyo ; A. O. Adewumi ; C. K. Ayo, "Stock price prediction using the arima model," in Proc. IEEE International Conference on Computer Modelling and Simulation (ICCMS), March 2014.
[7] S. Wichaidit ; S. Kittitornkun, "Stock price prediction using the arima model," in Proc. IEEE International Computer Science and Engineering Conference (ICSEC), November 2015, pp. 1-4.
[8] M. J. Kane, N. Price, M. Scotch, and P. Rabinowitz, "Comparison of arima and random forest time series models for prediction of avian infuenza h5n1 outbreaks," BMC bioinformatics, vol. 15, no. 1, p. 276, 2014.
[9] L. J. Cao and F. E. H. Tay, "Support vector machine with adaptive parameters in financial time series forecasting," IEEE Transactions on Neural Networks, vol. 14, no. 6, pp. 1506-1518, Nov 2003.
[10] Y. Lin ; H. Guo ; J. Hu, "An svm-based approach for stock market trend prediction," in Proc. IEEE International Joint Conference on Neural Networks (IJCNN), August 2013, pp. 1-7.
[11] G. C. Cawley and N. L. C. Talbot, "Over-fitting in model selection and subsequent selection bias in performance evaluation," Journal of Machine Learning Research, vol. 11, pp. 2079-2107, July 2010E.
[12] Y. B. Wijaya ; S. Kom ; T. A. Napitupulu, "Stock price prediction: Comparison of arima and artificial neural network methods - an indonesia stock's case," in Proc. IEEE International Conference on Advances in Computing, Control, and Telecommunication Technologies (AVT), December 2010, pp. 176-179.
[13] A. Graves ; A. Mohamed ; G. Hinton, "Speech recognition with deep recurrent neural networks," in Proc. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), May 2013, pp. 6645-6649.
[14] K. Dutta ; K. K. Sarma, "Multiple feature extraction for rnn-based assamese speech recognition for speech to text conversion application," in Proc. IEEE International Conference on Communications, Devices and Intelligent Systems (CODIS), Kolkata, India, December 2012.
[15] T. Gao ; Y. Chai ; Y. Liu, "Applying long short term memory neural networks for predicting stock closing price," in Proc. IEEE International Conference on Software Engineering and Service Science (ICSESS), Beijing, China, April 2017.
[16] G. Xavier ; A. Bordes ; Y. Bengio, "Deep sparse rectifier neural networks." in Proceedings of the fourteenth international conference on artificial intelligence and statistics (AISTATS), New York City, NY, USA, June 2011.
[17] T.Robert, "Regression shrinkage and selection via the lasso," Proceedings of the IEEE, vol. 58, no. 1, pp. 267-288, 1996.
[18] B. Leon, Large-scale machine learning with stochastic gradient descent. Springer, 2010, pp. 177-186.
[19] H. Kaiming, Z. Xiangyu, R. Shaoqing, and S. Jian, Delving deep into rectifiers: Surpassing human-level performance on imagenet classification, 2015, pp. 1026-1034.