這篇研究主要探討了機器學習在股票分析中的應用。本研究將目標放在理解如何利用機器學習找出最佳的股票分析方法組合，並評估這種方法是否能帶來優異的投資績效。研究首先進行了廣泛的文獻回顧，瞭解了各種與股票分析相關的研究，同時深入探討了將機器學習融入投資領域的各種設計與應用。緊接著，本研究收集和整理了台灣上市股的資料來作為實驗的樣本，並設計了一套雙階段的機器學習流程，也就是先篩選出每一季表現較好的個股作為當季的股票池，然後再基於它們的資料去預測每日的進出場點位。最終搭配普通買賣規則，即可進行完整的回測。研究結果顯示，本篇論文的研究方法無論是在篩選股票或是在預測進出場訊號上都能取得不錯的預測效果，同時也發現隨著訓練集資料的增加，預測準確度也能逐步得到提升。接著，本研究也針對進出場訊號和回測設定進行多種調整與檢驗，以了解不同情況下本策略能夠取得的投資績效。回測的結果顯示本策略可以根據投資者的風險偏好去進行不同的設定來放寬或限縮對於交易的限制，並以此來反映對於報酬和風險之間的抉擇。最後本研究也以表現較好的設定展示其回測績效，並得到了高於同時期的加權報酬指數將近一倍的夏普比率。

This study aims to understand how machine learning can be utilized to identify optimal stock analysis approaches and evaluate whether such methods can lead to superior investment performance. The study begins with examining various research related to stock analysis and exploring the applications of integrating machine learning into investment. Subsequently, this study collects and organizes data from stocks listed on the Taiwan Stock Exchange as samples. A two-stage machine learning process is designed, involving the selection of outperforming stocks as the quarterly stock pool, and predicting daily entry and exit points based on their data. The complete backtesting is conducted in conjunction with regular buy/sell rules. The research results demonstrate that the proposed methodology achieves favorable predictive accuracy in both stock selection and entry/exit signal forecasting. Furthermore, the study performs multiple adjustments on the entry/exit signals and backtesting settings to evaluate the investment performance under different scenarios. The backtesting results reveal that the strategy can be customized by adjusting trading restrictions based on investors' risk preferences, reflecting the trade-off between returns and risks. Finally, the study presents the backtesting performance using the well-performing settings and achieves a Sharpe ratio nearly twice as high as Taiwan Weighted Index during the same period.

論 文 審 定 書 ................................................................................................................... i 誌 謝 .................................................................................................................. ii 摘 要 ................................................................................................................. iii A b s t r a c t .................................................................................................................. iv 第一章 緒論 .................................................................................................................... 1
第一節 研究背景與動機............................................................................................. 1 第二節 研究目的......................................................................................................... 2 第三節 研究流程......................................................................................................... 2
第二章 文獻探討 ............................................................................................................ 4 第一節 股票分析方法................................................................................................. 4 第二節 機器學習與投資............................................................................................. 9
第三章 研究方法 .......................................................................................................... 11 第一節 樣本搜集與整理........................................................................................... 13 第二節 篩選股票....................................................................................................... 17 第三節 預測進出場訊號........................................................................................... 26 第四節 回測設定....................................................................................................... 33
第四章 實證結果 .......................................................................................................... 37 第一節 預測準確度................................................................................................... 37 第二節 回測結果....................................................................................................... 45 第三節 最佳結果....................................................................................................... 54
第五章 結論與建議 ...................................................................................................... 57 第一節 研究結論與貢獻........................................................................................... 57 第二節 研究建議與展望........................................................................................... 59
第六章 參考文獻 .......................................................................................................... 60 第一節 外文文獻....................................................................................................... 60 第二節 中文文獻....................................................................................................... 61
附 錄................................................................................................................. 62

Abarbanell, J. S., & Bushee, B. J. (1998). Abnormal Returns to a Fundamental Analysis Strategy. The Accounting Review, 73(1), 19-45.
Bettman, J. L., Sault, S. J., & Schultz, E. L. (2009). Fundamental and technical analysis: substitutes or complements? Accounting & Finance, 49(1), 21-36. https://doi.org/https://doi.org/10.1111/j.1467-629X.2008.00277.x
BLUME, L., EASLEY, D., & O'HARA, M. (1994). Market Statistics and Technical Analysis: The Role of Volume. The Journal of Finance, 49(1), 153-181. https://doi.org/https://doi.org/10.1111/j.1540-6261.1994.tb04424.x
FELTHAM, G. A., & OHLSON, J. A. (1995). Valuation and Clean Surplus Accounting for Operating and Financial Activities*. Contemporary Accounting Research, 11(2), 689-731. https://doi.org/https://doi.org/10.1111/j.1911-3846.1995.tb00462.x
Kuo, W.-L., Dai, T.-S., & Chang, W.-C. (2021). Solving Unconverged Learning of Pairs Trading Strategies with Representation Labeling Mechanism. CIKM Workshops,
Obthong, M., Tantisantiwong, N., Jeamwatthanachai, W., & Wills, G. (2020). A survey on machine learning for stock price prediction: algorithms and techniques 2nd International Conference on Finance, Economics, Management and IT Business (05/05/20 - 06/05/20)\, https://eprints.soton.ac.uk/437785/\
Park, C.-H., & Irwin, S. H. (2007). WHAT DO WE KNOW ABOUT THE PROFITABILITY OF TECHNICAL ANALYSIS? Journal of Economic Surveys, 21(4), 786-826. https://doi.org/https://doi.org/10.1111/j.1467-6419.2007.00519.x
Piotroski, J. D. (2000). Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers. Journal of Accounting Research, 38, 1-41. https://doi.org/10.2307/2672906
Tsai, C.-F., & P, W. (2009). Stock Price Forecasting by Hybrid Machine Learning Techniques. Lecture Notes in Engineering and Computer Science, 1.
Wafi, A. S., Hassan, H., & Mabrouk, A. (2015). Fundamental Analysis Models in Financial Markets – Review Study. Procedia Economics and Finance, 30, 939-947. https://doi.org/https://doi.org/10.1016/S2212-5671(15)01344-1
倪佳筠. (2023). 結合基本面與技術面之股票買賣策略-以台灣股市為例 淡江大學]. 新北市. https://hdl.handle.net/11296/8mkekn
吳威. (2022). 台股投資策略分析：籌碼面研究 輔仁大學]. 新北市. https://hdl.handle.net/11296/7767cb
李良俊. (2003). 台灣股票市場技術分析有效性之研究 實踐大學]. 台北市. https://hdl.handle.net/11296/4xf69p
李雅君. (2000). 資料擷取在財務報表上的運用：台灣地區電子上市公司報酬之預測 朝陽科技大學]. 台中市. https://hdl.handle.net/11296/k4gc97
林佳靜. (2019). 以改良F-score指標建構投資組合績效之探討. 管理資訊計算, 8, 1-23. https://doi.org/10.6285/MIC.201908/SP_01_8.0001 ,
柯靖家. (2023). 以機器學習模型預測台股大盤走勢 逢甲大學]. 台中市. https://hdl.handle.net/11296/5y69ww
楊蓁. (2022). 營收宣告的長短期間效果：公司財務比率與技術指標的角色 國立高雄科技大學]. 高雄市. https://hdl.handle.net/11296/cgk5p6
邱科毓. (2014). 均線分析投資績效探討：以臺灣50指數成份股為例 國立中山大學]. 高雄市. https://hdl.handle.net/11296/483v5y
陳彥安. (2019). 深度學習於台股動態投資組合之應用 國立交通大學]. 新竹市. https://hdl.handle.net/11296/86n2qg
陳浩瑋. (2017). 以深度學習進行股價預測之研究 國立高雄第一科技大學]. 高雄市. https://hdl.handle.net/11296/xa9tc2
黃離夏. (2022). 應用基本面分析與線性預測選擇最有潛力股票的框架 國立陽明交通大學]. 新竹市. https://hdl.handle.net/11296/ar95g2