隨著近年金融科技愈受歡迎及進步，愈來愈多新穎的數學方法及模型被應用於金融產業已用於增加獲利。在眾多方法中，以機器學習最為出名且經常用於相關文獻中以研究是否可以在最有效的時間內預測經濟及市場情勢。本文將利用相關特徵嘗試增進機器學習模型之精準度及準確度以預測臺灣市場狀況，再進一步結合Black-Litterman模型以建立有效且可獲利之投資組合。同時，針對Black-Litterman模型的觀點定義也將被進一步改進，使模型更加具有使用性。

研究結果顯示機器學習模型確實可為一預測市場狀況及標籤之有效方法，然而，資料庫之品質及量可能對精準度及準確度的影響極大，且當年分遭遇特殊事件時，精準度及準確率仍可維持並獲得報酬。

With the increasing popularity of financial technology (fintech), an increasing number of methods are used in the financial industry with the aim of improving profitability. Machine learning is the famous one that we often see in the research which try to predict the market’s situation. In this paper, machine learning algorithms will be utilized with Black-Litterman model. How to enhance the accuracy form predictions of machine learning and input the ideal parameters in the Black-Litterman model will be paper’s goal. The research will use machine learning, and add macroeconomic indicators and industry dummy as features to test if the machine learning can really find out better stocks in Taiwan and makes the new outperform portfolio. At the same time, the definition of the view in the Black-Litterman model will be improved in order to let the model be much more feasible.

As the results, machine learning can be a feasible tool in predict the labels of every stock. However, the quality and amount of data will affect the accuracy. Besides, when the specific incidents take place, accuracy and precision can be maintained and earn profits.

論文審定書 i
摘要 ii
Abstract iii
I. Introduction 1
1. General Background Information 1
2. Research Purpose 3
II. Literature Review 5
1. Black-Litterman Model 5
2. Factors in Model 7
3. Machine Learning 10
3.1 Machine Learning 10
3.2 Decision Tree 10
3.3 Feature Engineering and Feature Selection 11
4. Finance with Machine Learning 12
5. Performance Evaluation 14
5.1 Sharpe Ratio 15
5.2 Max Drawdown 16
5.3 Accuracy 16
5.4 Precision 17
III. Methodology 18
1. Data Description 19
1.1 Period of Data 19
1.2 Data Source 19
2. Tool 20
3. Black-Litterman Model 20
3.1 Covariance Matrix 20
3.2 Implied Excess Equilibrium Return 21
3.3 View of Investors 21
3.4 New Combined Return 23
4. Machine Learning 23
4.1 Feature Engineering 24
4.2 Classifier 26
4.3 Supervised Learning 28
4.4 Missing Data 30
4.5 Ensemble Model 30
5. Portfolio Construction 33
5.1 Portfolio Components 33
5.2 Weight 33
IV. Empirical Result 35
1. Benchmark 35
2. Descriptive Statistics 35
3. Results and Performances 35
3.1 Performance Analysis 36
3.2 Accuracy and Precision 39
3.3 Features Selection 44
4. Performance During Specific Period 48
4.1 Pandemic Crisis 48
4.2 Bullish Market After Bearish Market 56
V. Conclusion 61
References 66
Appendix 72

Figure
Figure 2- 1 The structure of new combined return forming 6
Figure 2- 2 Example of decision tree 11
Figure 3- 1 Research framework 18
Figure 3- 2 Training period example 19
Figure 4- 1Cumulative return of BL-Portfolio and benchmarks 36
Figure 4- 2 Cumulative return of Adj. BL-Portfolio and benchmarks 37
Figure 4- 3 Cumulative return of BL-Portfolio and benchmark during 2020 49
Figure 4- 4 Cumulative return of Adj. BL-Portfolio and benchmark during 2020 50
Figure 4- 5 Cumulative return of BL-Portfolio and benchmark during 2009 56
Figure 4- 6 Cumulative return of Adj. BL-Portfolio and benchmark during 2009 57

Table
Table 4- 1 Performance of BL-Portfolio and benchmark 36
Table 4- 2 Performance of the Adj. BL-Portfolio 37
Table 4- 3 Classifier A’s accuracy and precision 39
Table 4- 4 BL-Portfolio classifier A’s performance of prediction 40
Table 4- 5 Adj. BL-Portfolio classifier A’s performance of prediction 41
Table 4- 6 Classifier B’s accuracy and precision 41
Table 4- 7 Classifier B’s performance of prediction 42
Table 4- 8 Top 5 features* 44
Table 4- 9 Classifier A’s feature proportion 46
Table 4- 10 Classifier B’s feature proportion 47
Table 4- 11 Performance of Adj. BL-Portfolio and benchmarks during 2020 49
Table 4- 12 Performance of Adj. BL-Portfolio and benchmarks during 2020 50
Table 4- 13 Classifier A’s accuracy and precision during 2020 51
Table 4- 14 Classifier B’s accuracy and precision during 2020 51
Table 4- 15 Classifier A’s performance of prediction of portfolio during 2020 52
Table 4- 16 Classifiers’ performance of prediction during 2020 52
Table 4- 17 Classifier A’s feature proportion during 2020 54
Table 4- 18 Top 5 features during 2020* 55
Table 4- 19 Performance of BL-Portfolio and benchmark during 2009 57
Table 4- 20 Performance of BL-Portfolio and benchmark during 2009 58
Table 4- 21 Classifier A’s accuracy and precision during 2009 58
Table 4- 22 Classifier A’s performance of prediction of portfolio during 2009 59

English Literature
1. Ang, A., & Chen, J. (2002). Asymmetric correlations of equity portfolios. Journal of Financial Economics, 63(3), 443–494.
2. Assiri, A. S., Nazir, S., & Velastin, S. A. (2020). Breast Tumor Classification Us-ing an Ensemble Machine Learning Method. Journal of Imaging, 6(6), 39. https://doi.org/10.3390/jimaging6060039
3. Awan, F., Saleem, Y., Minerva, R., & Crespi, N. (2020). A Comparative Analysis of Machine/Deep Learning Models for Parking Space Availability Prediction. Sensors, 20. https://doi.org/10.3390/s20010322
4. Black, F., & Litterman, R. B. (1991). Asset Allocation: Combining Investor Views with Market Equilibrium. The Journal of Fixed Income. https://doi.org/10.3905/jfi.1991.408013
5. Breiman, L. (2001). Random Forests. Machine Learning, 45(1), 5–32. https://doi.org/10.1023/A:1010933404324
6. Breiman, L., Friedman, J., Stone, C. J., & Olshen, R. A. (1984). Classification and Regression Trees. Taylor & Francis.
7. Carhart, M. M. (1997). On Persistence in Mutual Fund Performance. The Journal of Finance, 52(1), 57–82. https://doi.org/10.2307/2329556
8. Chen, N.-F., Roll, R., & Ross, S. A. (1986). Economic Forces and the Stock Mar-ket. The Journal of Business, 59(3), 383–403.
9. Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794. https://doi.org/10.1145/2939672.2939785
10. Chlebus, M., & Osika, Z. (2020). Comparison of tree-based models performance in prediction of marketing campaign results using Explainable Artificial Intelli-gence tools (Working Paper No. 2020–15). Faculty of Economic Sciences, Uni-versity of Warsaw. https://econpapers.repec.org/paper/warwpaper/2020-15.htm
11. Daniel, K., Hirshleifer, D., & Sun, L. (2020). Short- and Long-Horizon Behavioral Factors. The Review of Financial Studies, 33(4), 1673–1736. https://doi.org/10.1093/rfs/hhz069
12. Daoud, E. A. (2019). Comparison between XGBoost, LightGBM and CatBoost Using a Home Credit Dataset. International Journal of Computer and Information Engineering, 13(1), 6–10.
13. Donthireddy, P. (2018). Black-Litterman Portfolios with Machine Learning de-rived Views. https://doi.org/10.13140/RG.2.2.26727.96160
14. Fama, E. F., & French, K. R. (1992). The Cross-Section of Expected Stock Re-turns. The Journal of Finance, 47(2), 427–465. https://doi.org/10.1111/j.1540-6261.1992.tb04398.x
15. Fama, E. F., & French, K. R. (2015). A five-factor asset pricing model. Journal of Financial Economics, 116(1), 1–22. https://doi.org/10.1016/j.jfineco.2014.10.010
16. G. Beat, B., C. Smith, R., & Ward, P. (2009). The Barra Europe equity model (EUE3). Research Notes.
17. Harris, R., Stoja, E., & Tan, L. (2016). The Dynamic Black-Litterman Approach to Asset Allocation. European Journal of Operational Research, 259. https://doi.org/10.1016/j.ejor.2016.11.045
18. Heaton, J. B., Polson, N. G., & Witte, J. H. (2016). Deep Learning in Finance. ArXiv. http://arxiv.org/abs/1602.06561
19. Hou, K., Xue, C., & Zhang, L. (2015). Digesting Anomalies: An Investment Ap-proach. The Review of Financial Studies, 28(3), 650–705. https://doi.org/10.1093/rfs/hhu068
20. Idzorek, T. (2007). A step-by-step guide to the Black-Litterman model: Incorpo-rating user-specified confidence levels. In S. Satchell (Ed.), Forecasting Expected Returns in the Financial Markets (pp. 17–38). Academic Press. https://doi.org/10.1016/B978-075068321-0.50003-0
21. Jhaveri, S., Khedkar, I., Kantharia, Y., & Jaswal, S. (2019). Success Prediction using Random Forest, CatBoost, XGBoost and AdaBoost for Kickstarter Cam-paigns. 2019 3rd International Conference on Computing Methodologies and Communication (ICCMC), 1170–1173. https://doi.org/10.1109/ICCMC.2019.8819828
22. Kang, X., Lin, G., Chen, Y., Zhao, F., Zhang, E., & Jing, C. (2020). Robust and secure zero-watermarking algorithm for color images based on majority voting pattern and hyper-chaotic encryption. Multimedia Tools and Applications, 79(1), 1169–1202. https://doi.org/10.1007/s11042-019-08191-y
23. Khanna, S. K. (2019). Machine Learning v/s Deep Learning. 06(02), 4.
24. Markowitz, H. (1952). Portfolio Selection. The Journal of Finance, 7(1), 77–91. https://doi.org/10.2307/2975974
25. Meucci, A. (2008). The Black-Litterman Approach: Original Model and Exten-sions (SSRN Scholarly Paper ID 1117574). Social Science Research Network. https://doi.org/10.2139/ssrn.1117574
26. Meucci, A. (2009). ‪Risk and asset allocation‬. Springer Science & Business Media. https://scholar.google.com/citations?view_op=view_citation&hl=en&user=9iZ0GUQAAAAJ&citation_for_view=9iZ0GUQAAAAJ:RHpTSmoSYBkC‬‬
27. Novy-Marx, R. (2013). The other side of value: The gross profitability premium. Journal of Financial Economics, 108(1), 1–28.
28. Prokhorenkova, L., Gusev, G., Vorobev, A., Dorogush, A. V., & Gulin, A. (2019). CatBoost: Unbiased boosting with categorical features. NeurIPS 2018 Paper. http://arxiv.org/abs/1706.09516
29. Quinlan, J. R. (1986). Induction of Decision Trees. Machine Language, 1(1), 81–106. https://doi.org/10.1023/A:1022643204877
30. Rasekhschaffe, K. C., & Jones, R. C. (2019). Machine Learning for Stock Selec-tion. Financial Analysts Journal, 75(3), 70–88. https://doi.org/10.1080/0015198X.2019.1596678
31. Samuel, A. L. (1959). Some Studies in Machine Learning Using the Game of Checkers. IBM Journal of Research and Development, 3(3), 210–229. https://doi.org/10.1147/rd.33.0210
32. Stambaugh, R., & Yuan, Y. (2017). Mispricing Factors. The Review of Financial Studies, 1270–1315. https://doi.org/10.1093/rfs/hhw107
33. Walters, J. (2008). The Black¬Litterman Model: A Detailed Exploration. 39.
34. Weiss, S. M., & Indurkhya, N. (1998). Predictive Data Mining: A Practical Guide. Morgan Kaufmann.
35. Yang, J., Qu, Z., & Liu, Z. (2014). Improved Feature-Selection Method Consider-ing the Imbalance Problem in Text Categorization. The Scientific World Journal, 2014, 1–17. https://doi.org/10.1155/2014/625342
36. Yang, Y.-Q. (2021). Empirical Study on Industrial Trend Factor Trading Strate-gies in Taiwan Stock Market. National Sun Yat-sen University.

中文文獻
1. 黃介琳, & 林朝陽. (2011). 考慮異質變異風險值估計之比較－以外匯美元為例. 貨幣觀測與信用評等, 63–71.