跳到主要內容

臺灣博碩士論文加值系統

(44.200.140.218) 您好!臺灣時間:2024/07/19 01:32
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:江宇祥
研究生(外文):Jian, Yu-Xiang
論文名稱:應用Q學習實現智慧交易代理人機制-以加密貨幣市場為例
論文名稱(外文):Q-learning Agent for Intelligent Trading in Cryptocurrency Market
指導教授:黃信嘉黃信嘉引用關係陳勤明陳勤明引用關係
指導教授(外文):Huang, Shin-JiaChen, Chin-Ming
口試委員:蘇玄啟周建新郭建良黃信嘉陳勤明
口試委員(外文):Su, Xuan-QiChou, Jian-HsinKuo, Chien-LiangHuang, Shin-JiaChen, Chin-Ming
口試日期:2022-07-18
學位類別:碩士
校院名稱:國立高雄科技大學
系所名稱:資訊財務碩士學位學程
學門:商業及管理學門
學類:財務金融學類
論文種類:學術論文
論文出版年:2022
畢業學年度:110
語文別:中文
論文頁數:41
中文關鍵詞:Q-學習加密貨幣
外文關鍵詞:Q-learningCryptocurrency
相關次數:
  • 被引用被引用:0
  • 點閱點閱:136
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
程式交易是根據事先定義的交易策略進行自動交易,然而,有學者指出事先定義交易策略的程式交易並不能對所有加密貨幣都產生獲利空間;而是要因應當前貨幣價格的變動趨勢來產生最適合的交易策略並產生獲利。因此,本研究將應用強化式學習之Q學習方法來建構智慧交易代理人,讓其能因應當前貨幣價格的變動趨勢來產生最適合的交易策略。再者,本研究也發現智慧代理人建構最適合交易策略的效能高低,是在於狀態(State)的描述,因此本研究也將提出狀態描述的方法並運用於加密貨幣市場中。本研究將使用6種加密貨幣,分別為BTC、ETH、VET、ADA、TRX和XRP,其中,BTC和ETH設定為上升趨勢,VET和ADA設定為盤整趨勢,TRX和XRP設定為下降趨勢。另外,本研究也將6種加密貨幣區分成3個時間區間,分別為5分鐘、15分鐘跟1小時,最後使用強化式學習之Q學習進行回測。實證結果顯示,在上升趨勢中,ETH在1小時區間內的年化報酬為725.48%,而在盤整趨勢中,VET在1小時區間內的年化報酬為-14.95%,最後在下降趨勢中,XRP在1小時區間內的年化報酬為-3.7%。若與買入並持有的策略進行比較,本研究發現不管是上升、盤整和下降趨勢,在1小時區間內,6種加密貨幣的年化報酬都會比買入並持有策略的年化報酬還要來得更好。
Program trading refers to automated trading based on predefined trading strategies. However, some scholars have pointed out that program trading with predefined trading strategies is unable to generate profitable opportunities for all cryptocurrencies; rather, it is necessary to generate optimal trading strategies and generate profits according to prevailing trends in currency prices. Therefore, this study applies the Q-learning method of reinforcement learning to construct intelligent trading agents that are capable of generating optimal trading strategies in response to the current trend of currency prices. In addition, this study identifies that the effectiveness of intelligent agents in constructing the optimal trading strategies is dependent on state descriptions. Therefore, this study also proposes a state description method and applies it to the cryptocurrency market. In this study, six cryptocurrencies, including BTC, ETH, VET, ADA, TRX, and XRP, were selected, where BTC and ETH were set as uptrends, VET and ADA were set as correction trends, and TRX and XRP were set as downtrends. Furthermore, the six cryptocurrencies were divided into three time intervals of 5 minutes, 15 minutes, and 1 hour, respectively, and then back testing was conducted by using Q-learning reinforcement learning. The results indicated that in the uptrend, ETH had an annualized return of 725.48% in the 1-hour interval, while in the consolidation trend, VET had an annualized return of -14.95% in the 1-hour interval, and finally, in the downtrend, XRP had an annualized return of -3.7% in the 1-hour interval. When comparing to a buy-and-hold strategies, this study found that the annualized returns of the six cryptocurrencies outperformed the buy-and-hold strategies in the 1-hour interval, regardless of the uptrend, consolidation and downtrend
摘要 i
Abstract ii
目錄 iv
表目錄 vi
圖目錄 vii
第一章、緒論 1
1.1 研究背景與動機 1
1.2 研究目的 2
1.3 研究架構 2
第二章、文獻探討 4
2.1 加密貨幣的背景 4
2.2 技術指標應用於股票市場 5
2.3 機器學習與深度學習應用於加密貨幣市場 6
2.4 強化學習應用於股票市場 7
第三章、研究方法 8
3.1 馬可夫決策 8
3.2 強化式學習 9
3.3 Q-learning 10

第四章、研究環境 13
4.1 研究流程 13
4.2 幣種資料挑選與資料切割 15
4.3 代理人的獎勵與動作定義 19
4.4 代理人的環境狀態定義 20
4.5 參數介紹 22
4.6 評估工具 23
第五章、實證結果 25
5.1 上升趨勢 25
5.2 盤整趨勢 29
5.3 下降趨勢 33
5.4 小結 37
第六章、結論與建議 38
6.1 結論小結 38
6.2 建議 38
6.3 研究限制 39
參考文獻 40


1.Ayala, J., García-Torres, M., Noguera, J. L. V., Gómez-Vela, F., & Divina, F. (2021). Technical analysis strategy optimization using a machine learning approach in stock market indices. Knowledge-Based Systems, 225, 107119.
2.Carta, S., Ferreira, A., Podda, A. S., Recupero, D. R., & Sanna, A. (2021). Multi-DQN: An ensemble of Deep Q-learning agents for stock market forecasting. Expert systems with applications, 164, 113820
3.Chakole, J. B., Kolhe, M. S., Mahapurush, G. D., Yadav, A., & Kurhekar, M. P. (2021). A Q-learning agent for automated trading in equity stock markets. Expert Systems with Applications, 163, 113761.
4.Chong, T. T. L., Ng, W. K., & Liew, V. K. S. (2014). Revisiting the Performance of MACD and RSI Oscillators. Journal of risk and financial management, 7(1), 1-12.
5.Greaves, A., & Au, B. (2015). Using the bitcoin transaction graph to predict the price of bitcoin.
6.Katsiampa, P., Corbet, S., & Lucey, B. (2019). Volatility spillover effects in leading cryptocurrencies: A BEKK-MGARCH analysis. Finance Research Letters, 29, 68-74.
7.Lahmiri, S., & Bekiros, S. (2019). Cryptocurrency forecasting with deep learning chaotic neural networks. Chaos, Solitons & Fractals, 118, 35-40.
8.McNally, S., Roche, J., & Caton, S. (2018, March). Predicting the price of bitcoin using machine learning. In 2018 26th euromicro international conference on parallel, distributed and network-based processing, 339-343.
9.Nakamoto, S. (2008). Bitcoin: A peer-to-peer electronic cash system. Decentralized Business Review, 21260.
10.Sun, X., Liu, M., & Sima, Z. (2020). A novel cryptocurrency price trend forecasting model based on LightGBM. Finance Research Letters, 32, 101084.
11.Urquhart, A. (2016). The inefficiency of Bitcoin. Economics Letters, 148, 80-82.

電子全文 電子全文(網際網路公開日期:20270710)
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top