透過您的圖書館登入
IP:18.216.114.23
  • 學位論文

以深度學習 LSTM 方法進行台灣加權股價指數預測

Analysis and prediction on Taiwan Capitalization Weighted Stock Index use deep learning LSTM method

指導教授 : 黃仕斌

摘要


技術分析一直在資本市場分析方法中佔有一席之地,而在現今人工智慧 蓬勃發展的時代背景下,使用機器學習方法進行財金相關應用已經成為一股 熱潮。 本文使用台灣加權股票指數共計十年日資料,使用開盤價、收盤價、最 低價、最高價與多種技術分析指標為樣本集,以前向切割、歸一化方法處理 資料,比較深度學習 LSTM、RNN、GRU 方法之混淆矩陣精確度,最終在分辨 上漲下跌方向上得到較好的結果。進而證明深度學習方法應用於台灣大盤股 票之可行性,探討不同深度學習算法應用於時間序列指數。

並列摘要


Technical analysis has always played a important role in analyzing the capital market. With the rapid development of artificial intelligence, using machine-learning methods for financial applications has become a hot trend. Using a decade of TAIEX as data, adopting a variety of technical analysis indicators as sample set in the research, I compared the precision of LSTM and RNN and GRUconfusion matrix in the deep- learning field by Walk Forward Test and normalization. Finally I get good results about determining direction of increase or decrease price. To prove that The feasibility of applying deep learning method to TAIEX,And discuss the application of different deep learning algorithms to time series.

參考文獻


參考文獻
〔1〕Jaspreet,"A Concise History of Neural Networks." From: https://towardsdatascience.com/a-concise-history-of-neural-networks- 2070655d3fec,2016
〔2〕LYNN,"耗時三十年,深度學習之父 Hinton 是怎麼讓一度衰頹的類神經網 路重迎曙光的呢?",From: https://kopu.chat/2017/11/03/dl-hinton/,2017
〔3〕G. E. Hinton* and R. R. Salakhutdinov,"Reducing the Dimensionality of Data with Neural Networks",Science ,2006
〔4〕Sepp Hochreiter,Jürgen Schmidhuber,"Long Short-Term Memory",The MIT PressJournals,ISSN: 0899-7667,2006

延伸閱讀