成功大學博碩士論文系統

簡易檢索 / 詳目顯示

回結果列表

研究生：	余東杰 Yu, Dong-Jie
論文名稱：	使用深度學習建立使用者選股與投資策略 Use deep learning to assist users in stock selection and investment strategies
指導教授：	王惠嘉 Wang, Hei-Chia
學位類別：	碩士 Master
系所名稱：	管理學院 - 工業與資訊管理學系碩士在職專班 Department of Industrial and Information Management (on the job class)
論文出版年：	2022
畢業學年度：	110
語文別：	中文
論文頁數：	52
中文關鍵詞：	長短期神經網路、股票、深度學習
外文關鍵詞：	LSTM, stock, deep learning
相關次數：	點閱：166 下載：52
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

金融市場一直是社會的重要經濟指標，股票市場為其中一項眾多民眾選擇的投資工具；其價格走勢的預測是一個相當熱門的議題，由於影響股市之因素多得難以準確的預測，許多計算方法與投資策略應用在股票市場趨勢預測，但仍難有明確的模型。本論文採用具有時序之深度學習預測台灣個股收盤漲跌趨勢，以上櫃上市公司做為研究，使用之資料集為10年歷史股票資訊；採用數個連續交易日來預測當日交易日的收盤漲與跌。研究中考慮的特徵有10項技術指標及三大法人買超與賣超，利用長短期神經網路深度學習使用移動窗格法進而預測當日收盤漲與跌；研究成果表明增加法人買超與賣超交易行為相較引用的論文中的結果來的優秀 2%，不僅有較好的準確率訓練過程也相對穩定。另外文中亦將台灣元大50 (0050) 投資與使用本研究模型進行挑選其表現優秀的成分股進行買賣交易；其結果表明使用本研究相較證券所公告台灣元大50 之 2013年至2019年報酬率12.58%，研究中模型策略報酬率15.51 % 相對優秀約3%，其中我們的主交易策略2015與2018投資效益更是沒有負面效益，顯得研究中策略更具優勢。

The financial market has always been an important economic indicator of society, and the stock market is one of the investment tools chosen by many people; the prediction of its price trend is a very popular topic. With investment strategies applied in stock market trend forecasting, it is still difficult to have a clear model. In this paper, deep learning with time series is used to predict the closing trend of individual stocks in Taiwan, and the OTC listed companies are used as research. The data set used is 10 years of historical stock information; several consecutive trading days are used to predict the closing up and down trends of the current trading day. fall. The characteristics considered in the research include 10 technical indicators and three major corporate over-buying and over-selling, using the long-term and short-term neural network deep learning to use the moving pane method to predict the closing up and down of the day; the research results show that increasing the number of corporate over-buying and overselling The hyper-trading behavior is 2% better than the results in the cited papers, not only has a better accuracy rate, but also the training process is relatively stable. In addition, the article will also invest in 0050 ETF and use this research model to select its outstanding constituent stocks for trading; the results show that the use of this research is compared with the stock exchange announcement of 0050 ETF from 2013 to 2019. The rate of return is 12.58%, and the rate of return of the model strategy in the study is 15.51%, which is relatively excellent and about 3%. Among them, our main trading strategies in 2015 and 2018 have no negative benefits, which shows that the strategy in the study has more advantages.

摘要................................................................................................................................ I
表目錄..........................................................................................................................XI
圖目錄........................................................................................................................ XII
第 1 章 緒論............................................................................................................1
1 研究背景與動機........................................................................................2
2 研究目的....................................................................................................3
3 研究範圍與限制........................................................................................4
4 研究流程....................................................................................................5
第 2 章 文獻探討....................................................................................................6
1 股票市場理論............................................................................................6
2 技術分析理論............................................................................................6
3 三大法人交易影響....................................................................................7
4 深度學習....................................................................................................8
5 小結............................................................................................................9
第 3 章 研究方法..................................................................................................10
1 問題描述..................................................................................................11
2 資料預處理..............................................................................................11
2.1 連續值輸入..................................................................................14
2.2 縮放值輸入..................................................................................17
2.3 離散值輸入..................................................................................18
2.4 法人買超與賣超..........................................................................19
2.5 交易稅與手續費..........................................................................20
3 深度學習..................................................................................................21
3.1 Recurrent neural network (RNN).................................................21
3.2 前向過程......................................................................................21
3.3 反向過程請參閱附錄 1： RNN 反向過程...............................23
3.4 激活函數(Activation Function)...................................................23
3.5 長期相依性(Long-Range Dependence)......................................24
4 長短期記憶..............................................................................................25
4.1 神經網路架構..............................................................................25
5 當日收盤漲與跌預測建置......................................................................26
6 模型評價指標..........................................................................................27
7 交易策略..................................................................................................29
8 小結..........................................................................................................30
第 4 章 實驗結果與討論......................................................................................31
1 系統環境建置..........................................................................................31
2 實驗方法..................................................................................................31
3 實驗結果..................................................................................................35
3.1 實驗一：開盤價、收盤價、最高價、最低價與指標比較......35
3.2 實驗二：法人買賣行為的影響..................................................36
3.3 實驗三：特徵權重效果的結果影響..........................................38
3.4 實驗四：策略買賣之年化報酬率..............................................40
第 5 章 結論及未來研究方向..............................................................................42
參考文獻......................................................................................................................43
附錄..............................................................................................................................45
附錄 1：RNN 反向過程：.........................................................................45
附錄 2：LSTM 前向過程與反向過程：...................................................47
附錄 3： LSTM Keras................................................................................51
                                    

英文
Chang, P.-C., Wu, J.-L., & Lin, J.-J. (2016). A Takagi–Sugeno fuzzy model combined with a support vector regression for stock trading forecasting. Applied soft computing, 38, 831-842.
Chung, H., & Shin, K.-s. (2020). Genetic algorithm-optimized multi-channel convolutional neural network for stock market prediction. Neural Computing and Applications, 32(12), 7897-7914.
Huang, S., & Liu, S. (2019). Machine learning on stock price movement forecast: the sample of the Taiwan stock exchange. International Journal of Economics and Financial Issues, 9(2), 189.
Kara, Y., Boyacioglu, M. A., & Baykan, Ö. K. (2011). Predicting direction of stock price index movement using artificial neural networks and support vector machines: The sample of the Istanbul Stock Exchange. Expert systems with applications, 38(5), 5311-5319.
Patel, J., Shah, S., Thakkar, P., & Kotecha, K. (2015). Predicting stock and stock price index movement using trend deterministic data preparation and machine learning techniques. Expert systems with applications, 42(1), 259-268.
Wei, J., Xu, Q., & He, C. (2022). Deep learning of predicting closing price through historical adjustment closing price. Procedia Computer Science, 202, 379-384.
Xiong, R., Nichols, E. P., & Shen, Y. (2015). Deep learning stock volatility with google domestic trends. arXiv preprint arXiv:1512.04916.
Zhong, X., & Enke, D. (2017). A comprehensive cluster and classification mining procedure for daily stock market return forecasting. Neurocomputing, 267, 152-168.

中文
許琮苓，基於長短期記憶網路模型之股市趨勢預測，國立成功大學工程科學研究所，碩士論文，中華民國 2019 年 7 月。
楊世璋，三大法人進出與台灣股票短期報酬關係之研究，PhD Thesis. National Central University，碩士論文，中華民國 2018 年 7 月。
李啟宏，法人買賣超佔成交量比例對股票報酬之影響，國立高雄科技大學，碩士論文，中華民國 2022 年 7 月。
神經網路中 epoch，iteration，batchsize 相關理解和說明.(2018). from https://www.itread01.com/content/1545933021.html
Cheers. (2019). 無「財」、無「才」可理？年輕人理財調查，財務自評不及格. fromhttps://web.cheers.com.tw/issue/2019/financial/article04.php
疫情影響新鮮人「凍薪」近 4 成企業調整徵才計畫. from https://udn.com/news/story/7269/4561413
商周財富網台灣的散戶，讓法人賺走多少錢？from https://udn.com/news/story/7269/4561413
行政院主計總處，108 年全年工業全年每人每月總薪資平均. from https://udn.com/news/story/7269/4561413

校內：立即公開
校外：立即公開

簡易檢索 / 詳目顯示

相關論文