本研究嘗利用保持資料的原始性與廣泛性，盡可能地納入所有與台灣加權指數相關的市場資料(日頻)，結合台灣與美國的總體經濟資料(月頻)，利用機森林演算法(Random Forest)對台灣加權指數的日資料進行預測。本研究所使用的資料有11個類別，經過資料的轉換與參數最佳化後得到共計711個變數，利用推進分析(Walk-Forward Analysis, WFA)的方式訓練19個隨機森林模型，分別對台灣加權指數漲跌做出預測，並使用整體學習(Ensemble Learning)中的Stacking方法，再一次訓練一個隨機森林模型，整合19個所得到的預測結果，得到最終對台灣加權指數的漲跌預測。最後設計一個簡單的投資策略在台指期進行投資，驗證此法於投資上的可行性，最後發現策略獲利因子2.3，平均每年的累計報酬37.5%，勝率65%。

This paper attempts to maintain the originality and breadth of data. I have incorporated as much as possible of all market data (on a daily basis) related to the Taiwan Capitalization Weighted Stock Index (TWII), and have combined the macroeconomic data of Taiwan and U.S. (on a monthly basis). Then I use these data to build the Random Forest model to forecast the daily direction of TWII. There are 11 categories of the data used in this paper. After data conversion and parameter optimization, a total of 711 variables were obtained. All 19 Random Forest models were trained and forecast the daily direction of TWII by means of Walk-Forward Analysis (WFA), respectively. Then I use Stacking method to build another new Random Forest model, which uses all these prediction results to train. Then I will get the final prediction results of the direction of TWII. At the end, a simple investment strategy is designed by the final prediction results for investing in Taiwan Stock Price Index Futures. Then I can know the feasibility of investment by the model prediction results. The strategy profit factor was found to be 2.3, with an Annualized Rate of Return of 37.5% and a winning percentage of 65%.

論文審定書 ⅰ
摘要 ii
Abstract iii
第一章 緒論 1
1.1研究動機與背景 1
第二章 文獻回顧 3
2.1決策樹(Decision Tree) 3
2.2集成學習(Ensemble Learning) 3
第三章 資料處理與介紹 6
3.1日頻資料 8
3.1.1技術指標 8
3.1.2台灣股價指數市場相關資料 23
3.2月頻資料 28
3.2.1台灣總體經濟資料 30
3.2.2美國總體經濟資料 32
3.2.3原物料 34
第四章 敘述統計 36
4.1台灣加權指數 36
第五章 研究方法 42
5.1機器學習流程 43
5.2隨機森林理論 44
5.3模型架構 48
5.4最佳化 54
5.4.1指標參數最佳化 55
5.5實證結果 60
5.5.1策略架構 60
5.5.2模型預測結果 62
5.5.2.1模型訓練結果 63
5.5.2.2策略績效與數據 68
第六章 結論與建議 73
參考文獻 75

1. Breiman, L. (2001). Random forests. Machine learning, 45(1), 5-32.
2. Díaz-Uriarte, R., & De Andres, S. A. (2006). Gene selection and classification of microarray data using random forest. BMC bioinformatics, 7(1), 3.
3. Freund, Y., & Schapire, R. E. (1997). A decision-theoretic generalization of on-line learning and an application to boosting. Journal of computer and system sciences, 55(1), 119-139.
4. Wolpert, D. H. (1992). Stacked generalization. Neural networks, 5(2), 241-259.
5. Zhou, Z. H. (2012). Ensemble methods: foundations and algorithms. Chapman and Hall/CRC.
6. Hall, L. O., Chawla, N., & Bowyer, K. W. (1998). Combining decision trees learned in parallel. In Working Notes of the KDD-97 Workshop on Distributed Data Mining (pp. 10-15).
7. Ho, T. K. (1995, August). Random decision forests. In Document analysis and recognition, 1995., proceedings of the third international conference on (Vol. 1, pp. 278-282). IEEE.
8. Pang, S., Kim, D., & Bang, S. Y. (2003). Membership authentication in the dynamic group by face classification using SVM ensemble. Pattern Recognition Letters, 24(1-3), 215-225.
9. Pardo, R. (1992). Design, testing, and optimization of trading systems (Vol. 2). John Wiley & Sons.
10. Belveal, L. D. (1985). Charting commodity market price behavior. Irwin Professional Publishing.
11. Dagnino, G. (2001). Profiting in bull or bear markets. McGraw-Hill.
12. Ratanapakorn, O., & Sharma, S. C. (2007). Dynamic analysis between the US stock returns and the macroeconomic variables. Applied Financial Economics, 17(5), 369-377.
13. Sirucek, M. (2012). Macroeconomic variables and stock market: US review.
14. Teresienė, D., Aarma, A., & Dubauskas, G. (2008). RELATIONSHIP BETWEEN STOCK MARKET AND MACROECONOMIC VOLATILITY. Transformations in Business & Economics, 7(14).
15. Sung, T. K., Chang, N., & Lee, G. (1999). Dynamics of modeling in data mining: interpretive approach to bankruptcy prediction. Journal of Management Information Systems, 16(1), 63-85.
16. Leigh, W., Purvis, R., & Ragusa, J. M. (2002). Forecasting the NYSE composite index with technical analysis, pattern recognizer, neural network, and genetic algorithm: a case study in romantic decision support. Decision support systems, 32(4), 361-377.
17. Garofalakis, M., Hyun, D., Rastogi, R., & Shim, K. (2003). Building decision trees with constraints. Data Mining and Knowledge Discovery, 7(2), 187-214.
18. Kass, G. V. (1980). An exploratory technique for investigating large quantities of categorical data. Applied statistics, 119-127.
19. Leo, B., Friedman, J. H., Olshen, R. A., & Stone, C. J. (1984). Classification and regression trees. Wadsworth International Group.
20. Quinlan, J. R. (1993). C4. 5: Programming for machine learning. Morgan Kauffmann, 38, 48.
21. Ting, K. M., & Low, B. T. (1996). Theory combination: an alternative to data combination.
22. Sias, R. W., Starks, L. T., & Titman, S. (2006). Changes in institutional ownership and stock returns: Assessment and methodology. The Journal of Business, 79(6), 2869-2910.
23. 林美珍, & 馬麗菁. (2002). 投資機構交易資訊與市場報酬之互動關係.
24. 宋嘉凌. (2007). 台灣股市與主要國際股市之相關性研究. 臺灣大學國際企業學研究所學位論文, 1-46.
25. 林玉樹. (2006). 總體經濟景氣循環變數與股市報酬之相關性-以台灣為例. 臺灣大學國際企業學研究所學位論文, 1-41.
26. 劉祥熹, & 涂登才. (2012). 美國股市及其總體經濟變數間關連性與波動性之研究─ VEC GJR DCC-GARCH-M 之模型應用. 經濟研究 (Taipei Economic Inquiry), 48(1), 139-189.