本篇文章使用強化學習與深度學習結合，打造股市自動交易系統。除了股市中的原始資料外，也加入了一些投資者常用的技術指標，給定前 10天的資料並使用全連接神經網路以及 Q 學習去訓練系統。
訓練系統時，分了兩組來訓練。第一組，把台灣 50 全部的成分股做為訓練資料，並測試其往後 2 年的表現；第二組，取台灣 50 中的 9 支電子股做為訓練資料，並測試其往後 2 年的表現。實驗結果顯示，第一組訓練成果與買入持有策略相比並無明顯差異，而第二組的表現明顯優於買入持有策略。
實驗結果證明，DQN 模型於特定情況下在股市自動交易系統會是有效的。

This article uses a combination of reinforcement learning and deep learning to create an automated stock trading system. In addition to the original data from the stock market, some technical indicators commonly used by investors are also added to the system.
When training the system, we divided it into two groups. In the first group, all constituent stocks of the Taiwan 50 were used as training data and their performance
was tested for the next 2 years. In the second group, 9 electronic stocks in the Taiwan 50 were used as training data and tested their performance for the next 2
years. The results show that there is no significant difference between the first group and the buy-and-hold strategy, while the second group significantly outperforms the buy-and-hold strategy.
The experimental results demonstrate that the DQN model is effective in certain situations in the automated stock trading system.

[1] D. Z. Anderson, C. Benkert, and D. D. Crouch. Neural Networks for Perception. Neural networks for perception /.
[2] Huang Guan Chi. Double q-network in automated stock trading. 2021.
[3] M. Lai. Playing atari with deep reinforcement learning.
[4] Jae Won Lee, Jonghun Park, O Jangmin, Jongwoo Lee, and Euyseok Hong. A multiagent approach to q-learning for daily stock trading. IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans, 37(6):864–877, 2007.
[5] Moshe Leshno, Vladimir Ya Lin, Allan Pinkus, and Shimon Schocken. Multilayer feedforward networks with a nonpolynomial activation function can approximate any function. Neural networks, 6(6):861–867, 1993.
[6] Timothy P Lillicrap, Jonathan J Hunt, Alexander Pritzel, Nicolas Heess, Tom Erez, Yuval Tassa, David Silver, and Daan Wierstra. Continuous control with deep reinforcement learning. arXiv preprint arXiv:1509.02971, 2015.
[7] Jerome H Saltzer, David P Reed, and David D Clark. End-to-end arguments in system design. ACM Transactions on Computer Systems (TOCS), 2(4):277–288, 1984.
[8] David Silver, Julian Schrittwieser, Karen Simonyan, Ioannis Antonoglou, Aja Huang, Arthur Guez, Thomas Hubert, Lucas Baker, Matthew Lai, Adrian Bolton, et al. Mastering the game of go without human knowledge. nature, 550(7676):354–359, 2017.
[9] Richard S Sutton, David McAllester, Satinder Singh, and Yishay Mansour. Policy gradient methods for reinforcement learning with function approximation. Advances in neural information processing systems, 12, 1999.
[10] Hado Van Hasselt, Arthur Guez, and David Silver. Deep reinforcement learning with double q-learning. In Proceedings of the AAAI conference on artificial intelligence, volume 30, 2016.
[11] Bayya Yegnanarayana. Artificial neural networks. PHI Learning Pvt. Ltd., 2009.
[12] Özal Yıldırım, Paweł Pławiak, Ru-San Tan, and U Rajendra Acharya. Arrhythmia detection using deep convolutional neural network with long duration ecg signals. Computers in biology and medicine, 102:411–420, 2018.