學術產出-Theses
Article View/Open
Publication Export
-
題名 投資組合管理：Black-Litterman模型結合不同機器學習方法
Portfolio Management: Black-Litterman Portfolios with Different Machine Learning Derived Views作者 李宏澤
Li, Hung-Tze貢獻者 蕭明福<br>廖四郎
Shaw, Ming-fu<br>Liao, Szu-Lang
李宏澤
Li, Hung-Tze關鍵詞 Black-Litterman模型
共變異數估計
機器學習模型
Black-Litterman model
Covariance matrix estimation
Machine learning日期 2023 上傳時間 10-Jul-2023 11:53:12 (UTC+8) 摘要 本研究嘗試以不同機器學習方法及不同預測目標，預測資產價格漲跌方向與幅度並結合Black-Litterman模型，建構全球化之投資組合資產配置。以金融資產之價量指標、技術指標及Fama-French三因子為輸入變數，在資料處理上避免使用KNN方式填補遺失值，確保資料的正確性。將機器學習模型預測結果代入Black-Litterman模型中的投資者觀點，結合不同共變異數估計方法，比較在不同投資策略下資產配置的績效表現。實證結果發現，Ledoit-Wolf Shrinkage Variance Estimate為最佳的共變異數估計方法，在分別預測價格漲跌與幅度時，XGBoost有較高的準確率；在直接預測價格漲跌與幅度時， Random Forest有較高的準確率；而在績效表現上，SVM模型於極大化夏普比率與超額報酬-風險值比率時，能有效地分散投資及降低風險；於測試集中，Random Forest直接預測價格漲跌與幅度的績效表現長時間優於其他模型，直到最後三個月，使用分別預測的方式能創造大量報酬，最後以XGBoost分別預測價格漲跌與幅度獲得最高的累積報酬率，並且超越iShares Russell 1000 ETF及直接預測價格漲跌與幅度的模型，造成模型表現差異的原因則源於模型組成與變數選擇。
This research attempts to use different machine learning methods and different forecasting objectives to predict the direction and volatility of asset price. Subsequently, combine the Black-Litterman model to construct a global portfolio asset allocation. Using the price and volume indicators of financial assets, technical indicators and the Fama-French three factors model as input variables. Additionally, avoid using the KNN method to fill in missing values in data processing to ensure the correctness of the data. Substitute the prediction results of the machine learning model into the investor`s point of view in the Black-Litterman model and combine different covariance estimation methods to compare the performance of asset allocation under different investment strategies.The empirical results show that Ledoit-Wolf shrinkage variance estimate is the best covariance estimation method. In addition, XGBoost has a higher accuracy rate in separately predicting the direction and volatility of price; Random Forest has a higher accuracy rate in direction predicting. In terms of performance, SVM model can effectively diversify investments and reduce risk when maximizing the Sharpe ratio and VaR. In test data, using Random Forest to predict the direction and volatility directly outperforms others for a long time. Until last three months, the way of predicting separately can generate large returns. Finally, XGBoost predicts separately has the highest final cumulative return, which even better than the iShares Russell 1000 ETF and the models which predict directly. The reason for the difference in model performance is due to the model composition and variable selection.參考文獻 [1] Ai, X. W., Hu, T., Li, X. & Xiong, H. (2010). Clustering High-frequency Stock Data for Trading Volatility Analysis. 2010 Ninth International Conference on Machine Learning and Applications, 333-338.[2] Beach, S.L. & Orlov, A.G. (2007). An application of the Black–Litterman model with EGARCH-M-derived views for international portfolio management. Fin Mkts Portfolio Mgmt, 21, 147–166.[3] Best, M. J. & Grauer, R. R. (1991). On the Sensitivity of Mean-Variance-Efficient Portfolios to Changes in Asset Means: Some Analytical and Computational Results. The Review of Financial Studies, 4(2), 315–342.[4] Black, F., & Litterman, R. (1991). Asset allocation: combining investor views with market equilibrium. The Journal of Fixed Income, 1(2), 7-18.[5] Black, F., & Litterman, R. (1992). Global portfolio optimization. Financial Analysts Journal, 48(5), 28-43.[6] Chen, T. & Guestrin, C. (2016). XGBoost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794.[7] Donthireddy, P. (2018). Black-Litterman portfolio with machine learning derived views. https://doi.org/10.13140/RG.2.2.26727.96160[8] Fama, E., & French, K. (2004). The Capital Asset Pricing Model: Theory and Evidence. Journal of Economic Perspectives. 18(3), 25-46.[9] Henrique, B. M., Sobreiro, V. A., & Kimura, H. (2019). Literature review: Machine learning techniques applied to financial market prediction. Expert Syst. Appl., 124, 226–251.[10] Markowitz, H. (1952). Portfolio Selection. Journal of Finance. 7(1), 77-99.[11] Meucci, A. (2010). The Black litterman Approach: Original Model and Extensions. Download from: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1117574.[12] Michaud, R. O. (1989). The Markowitz Optimization Enigma: Is Optimized Optimal? Financial Analysts Journal, 31-42.[13] Mossin, J. (1966). Equilibrium in a Capital Asset Market. Econometrical, 34(4), 768–783.[14] Ledoit O. & Wolf, M. (2004). A well-conditioned estimator for large-dimensional covariance matrices. J. Multivariate Anal. 88 (2), 365–4.[15] Ledoit O. & Wolf, M. (2021). Shrinkage estimation of large covariance matrices: Keep it simple, statistician? J. Multivariate Anal. 186.[16] Sharpe, W. F. (1964). Capital asset prices: A theory of market equilibrium under conditions of risk. Journal of Finance, 19 (3), 425-442.[17] Treynor, J. L. (1961). Market Value, Time, and Risk. Unpublished manuscript.[18] Treynor, J. L. (1962). Toward a Theory of Market Value of Risky Assets. Unpublished manuscript. A final version was published in 1999, in Asset Pricing and Portfolio Performance: Models, Strategy and Performance Metrics. Robert A. Krawczyk (editor) London: Risk Books, 15–22.[19] Zhang, C. & Tang, H. (2022). Empirical Research on Multifactor Quantitative Stock Selection Strategy Based on Machine Learning. 2022 3rd International Conference on Pattern Recognition and Machine Learning (PRML), 380-383.[20] Zhu, Y. (2021). Research on Financial Risk Control Algorithm Based on Machine Learning. 2021 3rd International Conference on Machine Learning, Big Data and Business Intelligence (MLBDBI), 16-19. 描述 碩士
國立政治大學
經濟學系
110258038資料來源 http://thesis.lib.nccu.edu.tw/record/#G0110258038 資料類型 thesis dc.contributor.advisor 蕭明福<br>廖四郎 zh_TW dc.contributor.advisor Shaw, Ming-fu<br>Liao, Szu-Lang en_US dc.contributor.author (Authors) 李宏澤 zh_TW dc.contributor.author (Authors) Li, Hung-Tze en_US dc.creator (作者) 李宏澤 zh_TW dc.creator (作者) Li, Hung-Tze en_US dc.date (日期) 2023 en_US dc.date.accessioned 10-Jul-2023 11:53:12 (UTC+8) - dc.date.available 10-Jul-2023 11:53:12 (UTC+8) - dc.date.issued (上傳時間) 10-Jul-2023 11:53:12 (UTC+8) - dc.identifier (Other Identifiers) G0110258038 en_US dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/145961 - dc.description (描述) 碩士 zh_TW dc.description (描述) 國立政治大學 zh_TW dc.description (描述) 經濟學系 zh_TW dc.description (描述) 110258038 zh_TW dc.description.abstract (摘要) 本研究嘗試以不同機器學習方法及不同預測目標，預測資產價格漲跌方向與幅度並結合Black-Litterman模型，建構全球化之投資組合資產配置。以金融資產之價量指標、技術指標及Fama-French三因子為輸入變數，在資料處理上避免使用KNN方式填補遺失值，確保資料的正確性。將機器學習模型預測結果代入Black-Litterman模型中的投資者觀點，結合不同共變異數估計方法，比較在不同投資策略下資產配置的績效表現。實證結果發現，Ledoit-Wolf Shrinkage Variance Estimate為最佳的共變異數估計方法，在分別預測價格漲跌與幅度時，XGBoost有較高的準確率；在直接預測價格漲跌與幅度時， Random Forest有較高的準確率；而在績效表現上，SVM模型於極大化夏普比率與超額報酬-風險值比率時，能有效地分散投資及降低風險；於測試集中，Random Forest直接預測價格漲跌與幅度的績效表現長時間優於其他模型，直到最後三個月，使用分別預測的方式能創造大量報酬，最後以XGBoost分別預測價格漲跌與幅度獲得最高的累積報酬率，並且超越iShares Russell 1000 ETF及直接預測價格漲跌與幅度的模型，造成模型表現差異的原因則源於模型組成與變數選擇。 zh_TW dc.description.abstract (摘要) This research attempts to use different machine learning methods and different forecasting objectives to predict the direction and volatility of asset price. Subsequently, combine the Black-Litterman model to construct a global portfolio asset allocation. Using the price and volume indicators of financial assets, technical indicators and the Fama-French three factors model as input variables. Additionally, avoid using the KNN method to fill in missing values in data processing to ensure the correctness of the data. Substitute the prediction results of the machine learning model into the investor`s point of view in the Black-Litterman model and combine different covariance estimation methods to compare the performance of asset allocation under different investment strategies.The empirical results show that Ledoit-Wolf shrinkage variance estimate is the best covariance estimation method. In addition, XGBoost has a higher accuracy rate in separately predicting the direction and volatility of price; Random Forest has a higher accuracy rate in direction predicting. In terms of performance, SVM model can effectively diversify investments and reduce risk when maximizing the Sharpe ratio and VaR. In test data, using Random Forest to predict the direction and volatility directly outperforms others for a long time. Until last three months, the way of predicting separately can generate large returns. Finally, XGBoost predicts separately has the highest final cumulative return, which even better than the iShares Russell 1000 ETF and the models which predict directly. The reason for the difference in model performance is due to the model composition and variable selection. en_US dc.description.tableofcontents 摘要 IAbstract II目次 III表次 IV圖次 V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第二章 文獻回顧 3第一節 機器學習方法相關文獻 3第二節 投資組合理論相關文獻 3第三章 研究方法 5第一節 研究對象 5第二節 Black-Litterman 模型 7第三節 Fama-French 三因子模型 9第四節 機器學習方法 9第五節 模型架構 13第四章 實證結果 21第一節 機器學習模型預測投資人觀點 21第二節 模型變數選擇 24第三節 模型績效表現 28第五章 結論與建議 36第一節 結論 36第二節 未來展望 36參考文獻 38附錄 41附錄一 Kenneth R. French Data Library 41附錄二 技術指標公式 41附錄三 EEM投資者觀點預測結果 43附錄四 EFA投資者觀點預測結果 44附錄五 GLD投資者觀點預測結果 45附錄六 IYR投資者觀點預測結果 46附錄七 TLT投資者觀點預測結果 47 zh_TW dc.format.extent 2854384 bytes - dc.format.mimetype application/pdf - dc.source.uri (資料來源) http://thesis.lib.nccu.edu.tw/record/#G0110258038 en_US dc.subject (關鍵詞) Black-Litterman模型 zh_TW dc.subject (關鍵詞) 共變異數估計 zh_TW dc.subject (關鍵詞) 機器學習模型 zh_TW dc.subject (關鍵詞) Black-Litterman model en_US dc.subject (關鍵詞) Covariance matrix estimation en_US dc.subject (關鍵詞) Machine learning en_US dc.title (題名) 投資組合管理：Black-Litterman模型結合不同機器學習方法 zh_TW dc.title (題名) Portfolio Management: Black-Litterman Portfolios with Different Machine Learning Derived Views en_US dc.type (資料類型) thesis en_US dc.relation.reference (參考文獻) [1] Ai, X. W., Hu, T., Li, X. & Xiong, H. (2010). Clustering High-frequency Stock Data for Trading Volatility Analysis. 2010 Ninth International Conference on Machine Learning and Applications, 333-338.[2] Beach, S.L. & Orlov, A.G. (2007). An application of the Black–Litterman model with EGARCH-M-derived views for international portfolio management. Fin Mkts Portfolio Mgmt, 21, 147–166.[3] Best, M. J. & Grauer, R. R. (1991). On the Sensitivity of Mean-Variance-Efficient Portfolios to Changes in Asset Means: Some Analytical and Computational Results. The Review of Financial Studies, 4(2), 315–342.[4] Black, F., & Litterman, R. (1991). Asset allocation: combining investor views with market equilibrium. The Journal of Fixed Income, 1(2), 7-18.[5] Black, F., & Litterman, R. (1992). Global portfolio optimization. Financial Analysts Journal, 48(5), 28-43.[6] Chen, T. & Guestrin, C. (2016). XGBoost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794.[7] Donthireddy, P. (2018). Black-Litterman portfolio with machine learning derived views. https://doi.org/10.13140/RG.2.2.26727.96160[8] Fama, E., & French, K. (2004). The Capital Asset Pricing Model: Theory and Evidence. Journal of Economic Perspectives. 18(3), 25-46.[9] Henrique, B. M., Sobreiro, V. A., & Kimura, H. (2019). Literature review: Machine learning techniques applied to financial market prediction. Expert Syst. Appl., 124, 226–251.[10] Markowitz, H. (1952). Portfolio Selection. Journal of Finance. 7(1), 77-99.[11] Meucci, A. (2010). The Black litterman Approach: Original Model and Extensions. Download from: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1117574.[12] Michaud, R. O. (1989). The Markowitz Optimization Enigma: Is Optimized Optimal? Financial Analysts Journal, 31-42.[13] Mossin, J. (1966). Equilibrium in a Capital Asset Market. Econometrical, 34(4), 768–783.[14] Ledoit O. & Wolf, M. (2004). A well-conditioned estimator for large-dimensional covariance matrices. J. Multivariate Anal. 88 (2), 365–4.[15] Ledoit O. & Wolf, M. (2021). Shrinkage estimation of large covariance matrices: Keep it simple, statistician? J. Multivariate Anal. 186.[16] Sharpe, W. F. (1964). Capital asset prices: A theory of market equilibrium under conditions of risk. Journal of Finance, 19 (3), 425-442.[17] Treynor, J. L. (1961). Market Value, Time, and Risk. Unpublished manuscript.[18] Treynor, J. L. (1962). Toward a Theory of Market Value of Risky Assets. Unpublished manuscript. A final version was published in 1999, in Asset Pricing and Portfolio Performance: Models, Strategy and Performance Metrics. Robert A. Krawczyk (editor) London: Risk Books, 15–22.[19] Zhang, C. & Tang, H. (2022). Empirical Research on Multifactor Quantitative Stock Selection Strategy Based on Machine Learning. 2022 3rd International Conference on Pattern Recognition and Machine Learning (PRML), 380-383.[20] Zhu, Y. (2021). Research on Financial Risk Control Algorithm Based on Machine Learning. 2021 3rd International Conference on Machine Learning, Big Data and Business Intelligence (MLBDBI), 16-19. zh_TW
