股票交易市場係針對已發行之股票進行買賣、轉讓與流通的場所，在現今資本集中的社會中已是一個不可或缺的制度。股份公司發起人為了完成公司設立、公司決策者為了集眾人之力，往往會透過公開發行股票向公眾募集資本以用於公司營運，這些募集而來的資本分為股票交由應募人或買進股票人持有，此稱為股東，股東享有公司資產與營運收益的權利。
然而，股票的價值有其變動性，並非歸於一定，可能會隨著公司的業績表現、盈利狀況、總體市場變化，甚至是政治、媒體、公眾情緒而波動，這表示各該股東所持有的股票之價值是變動的，且沒有精準的公式可供計算，因此很難精準預測該如何處理手中持有的股票。雖然股票市場的起伏以及股票交易價格的變動是難以預測且不停變化，但歷年來已有許多研究希望可以藉由某種方法預測股價變化以達到一定精準度的未來股價判斷。因為股票交易價格的變動會使股票持有人在不同時間點買進、賣出產生價差，而這些價差的存在即為短期股票交易者獲利的入手處。如果改善預測股價的方法，使我們能夠精準預測股票交易價格的變化，就可以做到在低價買入，待到高價賣出，以此來賺取股票交易的價差。
股票的價格可以視為一個時序的訊號，以下簡稱股票訊號。股票訊號是一維非線性且時變的系統。在預測非線性的訊號時，人工智慧領域裡的機器學習對於非線性的模型表現十分優秀。其中卷積神經網路的長處是特徵抽取，而疊層長短期記憶神經網路的優點是時序記憶，本篇論文希望可以結合這兩個神經網路的優點，並用來預測股票訊號。然而研究成果指出直接結合兩個神經網路，其表現比單純使用疊層長短期記憶神經網路還差。本論文其中一項亮點就在於結合兩者的「短路連結架構」，能夠成功合併兩者優點。使用短路連結的架構不只解決退化問題，使結合後的模型表現的比單純使用疊層長短期記憶神經網路更為精準。

The stock market is an indispensable element in this capital-intensive society. In order to obtain more capital to run the company or implement projects, newly-founded companies often use stocks as a means of raising funds. The value of each stock is fluctuating rather than a fixed price. The stocks of different companies will have different values because of the company's performance. Because the prices of stocks represent traders' expectations for the company's future, there is no fixed formula for stock changes, and it is difficult to accurately predict.
Although the changes in stock trading prices are difficult to predict and constantly changing, many studies still hope that a certain degree of accuracy can be achieved by some predict method. Because changes in stock trading prices will cause different prices at different time. If you can predict the ups and downs of stock trading prices, you can buy at a lower price and sell at a higher price to earn the spread.
The stock signal is a one-dimensional nonlinear and time-varying system. When predicting nonlinear signals, machine learning in the field of artificial intelligence performs well on nonlinear models. The advantage of convolutional neural networks is feature extraction, while the advantage of long short term memory neural networks is temporal memory. This paper hopes to combine the advantages of these two neural networks and use them to predict stock prices.
However, research results indicate that directly combining two neural networks performs worse than simply using a long short term memory neural network. One of the highlights of this paper is the combination of the "shortcut connection" of the two, which can successfully merge the advantages of the two neural network.

[1] Ritika Singh and Shashi Srivastava, “Stock prediction using deep learning,” Multimed Tools Appl, p. 18569–18584, 2016.
[2] Erkam Guresen, Gulgun Kayakutlu and Tugrul U. Daim, “Using artificial neural network models in stock market index prediction,” Expert Systems with Applications, pp. 10389-10397, 2011.
[3] A. M. TURING, “I.—COMPUTING MACHINERY AND INTELLIGENCE,” Mind, p. 433–460, 1950.
[4] Shu-Hsien Liao, “Expert system methodologies and applications—a decade review from 1995 to 2004,” Expert Systems with Applications, pp. 93-103, 2005.
[5] Sreelekshmy Selvin, Vinayakumar R, Gopalakrishnan E.A and Vijay Krishna Menon, Soman K.P, “STOCK PRICE PREDICTION USING LSTM,RNN AND CNN-SLIDING WINDOW MODEL,” International Conference on Advances in Computing, Communications and Informatics (ICACCI), pp. 1643-1647, 2017.
[6] Khaled A. Althelaya, El-Sayed M. El-Alfy and Salahadin Mohammed, “Stock Market Forecast Using Multivariate Analysis with Bidirectional and Stacked (LSTM, GRU),” 21st Saudi Computer Society National Computer Conference (NCC), 2018.
[7] J. Eapen, D. Bein and A. Verma, “Novel Deep Learning Model with CNN and Bi-Directional LSTM for Improved Stock Market Index Prediction,” IEEE 9th Annual Computing and Communication Workshop and Conference (CCWC), pp. 0264-0270, 2019.
[8] Kaiming He, Xiangyu Zhang, Shaoqing Ren and Jian Sun, “Deep Residual Learning for Image Recognition,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770-778, 2016.
[9] I.A Basheer and M Hajmeer, “Artificial neural networks: fundamentals, computing, design, and application,” Journal of Microbiological Methods, pp. 3-31, 2000.
[10] Rumelhart, D., Hinton, G. and Williams, R., “Learning representations by back-propagating errors,” Nature, p. 533–536, 09 October 1986.
[11] GuoqiangZhang,B.Eddy Patuwo and MichaelY. Hu, “Forecasting with artificial neural networks:: The state of the art,” International Journal of Forecasting, pp. 35-62, 1998.
[12] A. Sherstinsky, “Fundamentals of Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) network,” Physica D: Nonlinear Phenomena, p. 132306, 2020.
[13] M. Schuster and K. K. Paliwal, “Bidirectional recurrent neural networks,” IEEE Transactions on Signal Processing, pp. 2673-2681, Nov 1997.
[14] S. Hochreiter and J. Schmidhuber, "Long Short-Term Memory," Neural Computation, vol. 9, no. 8,, pp. 1735-1780, 1997.
[15] O'Shea, Keiron and Ryan Nash, “An introduction to convolutional neural networks,” arXiv preprint arXiv, 2015.