本研究因為以台灣上市櫃公司為股票池，建立準被動防禦性投資組合。首先以 Barra EUE3 的九個風格因子為基底，以 Long 五分位距的方式建立單因子投資組合，接著以風險平價方式，計算每個因子投組的權重，形成多因子投資組合。因為考量到，風險平價投組的限制。本研究引入 Black-Litterman 框架，使投資組合能根據投資人的觀點動態調整。而投資人的主動觀點，則以橫斷面報酬估計，並且利用因子擇時，以機器學習(隨機森林、XGBoost)預測 IC 值來判定是否要加入該因子的橫斷面報酬，最終以（Markowitz 1952）的最佳投資組合理論解得風險平價結合 Black-Litterman 投資組合的權重，以進行績效回測。經過實證發現(1)IC 值顯著異於零的因子(Momentum、Growth)在五爪圖與預測時都有較好的效 果。(2)XGBoost 相較隨機森林有更高的準確率，且更高的投組報酬。(3)風險平價 投資組合比台灣加權指數有更高夏普比率。(4)在績效方面， BL-RandomForest、 BL- XGBoost 的年化報酬、投資組合波動度、夏普比率、最大回測，皆優於台灣 加權指數。(BL-RandomForest、BL- XGBoost 投資組合為風險平價多因子投組當 作先驗報酬，以 RandomForest、XGBoost 演算法進行因子擇時)。以實務面來說，此種投資組合可以稱為 Smart Beta。代表著主動投資，被動管理。透過因子的特 性去形成投資組合，接者以投資經理人的角度去調整各個因子的權重，藉以此來 優化投資組合。總結來說，本研究將許多要素加入模型，以此優化投資組合。

In this study, we use the stock list which comes from Taiwan Stock Exchange (TSE) as our universe. At first, referring to Barra EUE3 9 style factors are base. each single factor portfolio is constructed by Long-short quintiles. Next, we combine nine single portfolios into a multi-factor portfolio by risky parity. Given the limits of risky parity, in this study, we add the framework of Black-Litterman models. With this method, our portfolio can dynamically adjust by investors’ views. As for investors’ views, we use the cross-sectional return to estimate. In addition, we adopt machine learning to forecast information ratio (IC) that can help us to select good factors in cross-section. Finally, we use (Markowitz 1952) mean-variance method to calculate the portfolio’s weight which is constructed by risk parity, Marching learning for factor timing, and Black-Litterman Models. Empirical analyses that the factors of IC which significantly different from zero have good performance in machine learning predict.
Furthermore, compared to RandomForest, XGBoost have higher accuracy. Besides, risk parity portfolio’s Sharpe ratio is higher than Taiwan SE weighted index. Lastly, the performance of BL- RandomForest, BL- XGBoost is superior to TWSE including (Return, Volatility, Sharpe ratio, Max drawdown ). In summary, we combine lots of ingredient into these models in order to optimize portfolio.

論文審定書 .................................................................................................................................... i
摘要 .. ….. ...................................................................................................................................... ii
Abstract ......................................................................................................................................... iii
目 錄 ………………………………………… ........................................................................... iv
圖目錄 .......................................................................................................................................... vi
表目錄 ......................................................................................................................................... vii
第一章 緒論 .................................................................................................................................. 1
1-1. 研究動機 ...................................................................................................................... 1
1-2. 研究目的 ...................................................................................................................... 1
1-3. 研究架構 ...................................................................................................................... 2
第二章 文獻回顧 .......................................................................................................................... 3
2-1. 多因子模型與 Black-Litterman .................................................................................. 3
2-2. 風險平價 ...................................................................................................................... 4
2-3. 因子擇時 ...................................................................................................................... 4
2-4. 機器學習於財務領域應用 .......................................................................................... 5
第三章 研究方法 .......................................................................................................................... 6
3-1. 研究流程 ...................................................................................................................... 6
3-2. 研究資料 ...................................................................................................................... 8
3-3. 因子建構 ...................................................................................................................... 9
3-3-1. 要素與因子 ............................................................................................................. 9
3-3-2. 因子投組建構 ....................................................................................................... 10
3-3-3. 橫斷面因子模型 ................................................................................................... 10
3-4. 風險平價 .................................................................................................................... 11
3-5. 因子擇時 .................................................................................................................... 12
3-6. XgBoost & Randomforest演算法 ............................................................................. 13
3-7. 資產配置 .................................................................................................................... 15
3-7-1 Blcak-Litterman介紹 .................................................................................................... 15
3-7-2參數輸入 ........................................................................................................................ 16
3-7-3投組建構 ........................................................................................................................ 18
3-8. 回測績效 .................................................................................................................... 20
第四章 實證結果 ........................................................................................................................ 22
4-1. 因子擇時模型輸出及輸入變項敘述統計 ................................................................ 22
4-2. 機器學習預測準確率 ................................................................................................ 26
4-3. 投資人觀點預期報酬分析 ........................................................................................ 27
4-4. 投資組合績效評比 .................................................................................................... 28
4-5. 模型績效評比 ............................................................................................................ 29
第五章 研究結論與建議 ............................................................................................................ 31
第六章 參考文獻 ........................................................................................................................ 32
附錄一、要素與因子定義 ......................................................................................................... 33
附錄二、因子投組橫斷面曝險 ................................................................................................. 37

Arnott, R. D., Clements, M., Kalesnik, V., & Linnainmaa, J. T. (2021). Factor
momentum. Available at SSRN 3116974.
Asness, C. S., Frazzini, A., & Pedersen, L. H. (2012). Leverage aversion and risk parity.
Financial analysts journal, 68(1), 47-59.
Asness, C. S., Friedman, J. A., Krail, R. J., & Liew, J. M. (2000). Style timing: Value
versus growth. The Journal of Portfolio Management, 26(3), 50-60.
Biner, B., Smith, R. C., & Ward, P. (2009). The Barra Europe Equity Model (EUE3).
MSCI Research Notes.
Black, F., & Litterman, R. (1990). Asset allocation: combining investor views with
market equilibrium. Goldman Sachs Fixed Income Research, 115.
Black, F., & Litterman, R. (1991). Global asset allocation with equities, bonds, and
currencies. Fixed Income Research, 2(15-28), 218.
Black, F., & Litterman, R. (1992). Global portfolio optimization. Financial analysts
journal, 48(5), 28-43.
Chaves, D., Hsu, J., Li, F., & Shakernia, O. (2011). Risk parity portfolio vs. other asset
allocation heuristic portfolios. The Journal of Investing, 20(1), 108-118.
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,
Costa, G., & Kwon, R. H. (2019). Risk parity portfolio optimization under a markov
regime-switching framework. Quantitative Finance, 19(3), 453-471.
Figelman, I. (2017). Black–Litterman with a Factor Structure Applied to Multi-Asset
Portfolios. The Journal of Portfolio Management, 44(2), 136-155.
Graham, B., Dodd, D. L. F., & Cottle, S. (1934). Security analysis (Vol. 452). McGraw-
Hill New York.
Grinold, R. C., & Kahn, R. N. (2000). Active portfolio management.
Haddad, V., Kozak, S., & Santosh, S. (2020). Factor timing. The Review of Financial Studies, 33(5), 1980-2018.
Hodges, P., Hogan, K., Peterson, J. R., & Ang, A. (2017). Factor timing with crosssectional
and time-series predictors. The Journal of Portfolio Management,
44(1), 30-43.
Jidong, L., & Ran, Z. (2018). Dynamic weighting multi factor stock selection strategy
based on XGboost machine learning algorithm. 2018 IEEE International
Conference of Safety Produce Informatization (IICSPI),
Johansen, S., & Nielsen, B. (2013). Outlier detection in regression using an iterated onestep
approximation to the Huber-skip estimator. Econometrics, 1(1), 53-70.
Markowitz, H. (1952). Portfolio Selection, Journal of Finance, March.
Min, L., Dong, J., Liu, D., & Kong, X. (2021). A black-litterman portfolio selection
model with investor opinions generating from machine learning algorithms.
Engineering Letters, 29(2), 710-721.
Rasekhschaffe, K. C., & Jones, R. C. (2019). Machine learning for stock selection.
Financial analysts journal, 75(3), 70-88.
Spinu, F. (2013). An algorithm for computing risk parity weights. Available at SSRN
2297383.