Publications-Theses
Article View/Open
Publication Export
-
題名 情緒分析交易策略設計
Formulate a Trading Strategy Using Sentiment Analysis作者 吳文萱
Wu, Wen-Xuan貢獻者 謝明華
Hsieh, Ming-Hua
吳文萱
Wu, Wen-Xuan關鍵詞 機器學習
情緒分析
新聞情緒
交易策略
Machine learning
Sentiment analysis
News sentiment
Trading strategy日期 2019 上傳時間 5-Sep-2019 15:48:06 (UTC+8) 摘要 過去的文獻提供了財經新聞情緒、社群新聞情緒等對金融商品價格之間具有顯著相關性性的信心實證。Zhang and Skiena (2009) 中使用由大規模自然語言處理（Natural Language Processing , NLP）新聞分析系統生成的新聞數據，對公司的新聞發布頻率，情緒和主觀性如何預測或反應在其股票交易量和報酬上進行研究，並提供基於新聞市場的交易策略。Feuerriegel and Prendinger (2016) 中則肯定新聞情緒能解釋股價走勢並以新聞文本分析建構交易策略。本實證研究基於欲探討情緒分析對金融資產走勢是否具有影響性及預測力，利用數種監督式之機器學習分類方法，包括邏輯斯回歸 (Logistic Regression)、隨機森林 (Random Forest)、梯度提升 (Gradient Boosting)、自適應提升(Adaptive Boosting) 以及支持向量機 (Support Vector Machine)，以台灣指數期貨的未來漲跌作為模型預測目標，尋找其中預測力最佳之模型並以此建構台指期交易策略。實證結果發現加入新聞情緒變數能有效提升多數模型之預測力，本實證研究挑選當日及隔日漲跌預測最佳之模型建構交易策略，在測試集之表現皆能擊敗買進持有策略及動能投資策略，預測隔日漲跌之策略表現優於預測當日漲跌之策略，情緒資料具有延遲顯現的效果，且其中以梯度提升模型預測之策略表現最佳。
Studies in the past provides evidence of confidence in connection with news sentiment and financial assets trend. Zhang and Skiena (2009) uses news data including the companys’ news release frequency and sentiment generated by a large-scale Natural Language Processing (NLP) news analysis system to predict or reflect on its stock return and trading volume. The study also formulate trading strategies based on the news market. Feuerriegel and Prendinger (2016) confirmed that news sentiment can explain stock price movements and construct trading strategies with news text mining.The empirical study is also based on the need to explore whether sentiment analysis has an impact and predictive power on financial asset trends. The empirical study uses several supervised machine learning classification methods, including logistic regression, random forest, gradient boosting, adaptive boosting, and support vector machine. I predict the future rise and fall of Taiwan index futures and look for the model with the best predictive power to construct the trading strategy.The empirical result shows that the addition of news sentiment variables can effectively improve the predictive power of all models. This empirical study selects the best model to construc trading strategy with the best forecast of the day and the next day. The performance of the test set can beat the buy-and-hold strategy and momentum investment strategy and the strategy predicts that the strategy of the next day`s change is better than the forecast of the day`s change. The news sentiment has the effect of delaying the emergence, and the strategy of using the gradient boosting model is outperforming.參考文獻 Antweiler, W., & Frank, M. Z. (2004). Is all that talk just noise? The information content of internet stock message boards. The journal of Finance, 59(3), 1259-1294.Bellman, R., & Lee, E. S. (1978). Functional equations in dynamic programming. Aequationes Mathematicae, 17(1), 1-18.Breiman, L., Friedman, J., Olshen, R., & Stone, C. (1984). Classification and regression trees. Wadsworth Int. Group, 37(15), 237-251.Chan, W. S. (2003). Stock price reaction to news and no-news: drift and reversal after headlines. Journal of Financial Economics, 70(2), 223-260.Cortes, C., & Vapnik, V. (1995). Support-vector networks. Machine learning, 20(3), 273-297.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.Friedman, J. H. (2001). Greedy function approximation: a gradient boosting machine. Annals of statistics, 1189-1232.Feuerriegel, S., & Prendinger, H. (2016). News-based trading strategies. Decision Support Systems, 90, 65-74.Haugeland, J. (1985). Artificial intelligence: the very idea. In: Cambridge, MA: MIT Press.Ho, T. K. (1995). Random decision forests. Paper presented at the Proceedings of 3rd international conference on document analysis and recognition.Keating, C., & Shadwick, W. F. (2002). A universal performance measure. Journal of performance measurement, 6(3), 59-84.Machinery, C. (1950). Computing machinery and intelligence-AM Turing. Mind, 59(236), 433.Malkiel, B. G., & Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. The journal of Finance, 25(2), 383-417.Shannon, C. E. (1948). A mathematical theory of communication. Bell system technical journal, 27(3), 379-423.Sharpe, W. F. (1994). The sharpe ratio. Journal of portfolio management, 21(1), 49-58.Sortino, F. A., & Price, L. N. (1994). Performance measurement in a downside risk framework. the Journal of Investing, 3(3), 59-64.Sul, H., Dennis, A. R., & Yuan, L. I. (2014). Trading on twitter: The financial information content of emotion in social media. Paper presented at the 2014 47th Hawaii International Conference on System Sciences.Tetlock, P. C., Saar‐Tsechansky, M., & Macskassy, S. (2008). More than words: Quantifying language to measure firms` fundamentals. The journal of Finance, 63(3), 1437-1467.Wilder, J. W. (1978). New concepts in technical trading systems: Trend Research.Young, T. W. (1991). Calmar ratio: A smoother tool. Futures, 20(1), 40.Zhang, W., & Skiena, S. (2009). Trading strategies to exploit news sentiment. Submitted for publication. 描述 碩士
國立政治大學
風險管理與保險學系
106358006資料來源 http://thesis.lib.nccu.edu.tw/record/#G0106358006 資料類型 thesis dc.contributor.advisor 謝明華 zh_TW dc.contributor.advisor Hsieh, Ming-Hua en_US dc.contributor.author (Authors) 吳文萱 zh_TW dc.contributor.author (Authors) Wu, Wen-Xuan en_US dc.creator (作者) 吳文萱 zh_TW dc.creator (作者) Wu, Wen-Xuan en_US dc.date (日期) 2019 en_US dc.date.accessioned 5-Sep-2019 15:48:06 (UTC+8) - dc.date.available 5-Sep-2019 15:48:06 (UTC+8) - dc.date.issued (上傳時間) 5-Sep-2019 15:48:06 (UTC+8) - dc.identifier (Other Identifiers) G0106358006 en_US dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/125540 - dc.description (描述) 碩士 zh_TW dc.description (描述) 國立政治大學 zh_TW dc.description (描述) 風險管理與保險學系 zh_TW dc.description (描述) 106358006 zh_TW dc.description.abstract (摘要) 過去的文獻提供了財經新聞情緒、社群新聞情緒等對金融商品價格之間具有顯著相關性性的信心實證。Zhang and Skiena (2009) 中使用由大規模自然語言處理（Natural Language Processing , NLP）新聞分析系統生成的新聞數據，對公司的新聞發布頻率，情緒和主觀性如何預測或反應在其股票交易量和報酬上進行研究，並提供基於新聞市場的交易策略。Feuerriegel and Prendinger (2016) 中則肯定新聞情緒能解釋股價走勢並以新聞文本分析建構交易策略。本實證研究基於欲探討情緒分析對金融資產走勢是否具有影響性及預測力，利用數種監督式之機器學習分類方法，包括邏輯斯回歸 (Logistic Regression)、隨機森林 (Random Forest)、梯度提升 (Gradient Boosting)、自適應提升(Adaptive Boosting) 以及支持向量機 (Support Vector Machine)，以台灣指數期貨的未來漲跌作為模型預測目標，尋找其中預測力最佳之模型並以此建構台指期交易策略。實證結果發現加入新聞情緒變數能有效提升多數模型之預測力，本實證研究挑選當日及隔日漲跌預測最佳之模型建構交易策略，在測試集之表現皆能擊敗買進持有策略及動能投資策略，預測隔日漲跌之策略表現優於預測當日漲跌之策略，情緒資料具有延遲顯現的效果，且其中以梯度提升模型預測之策略表現最佳。 zh_TW dc.description.abstract (摘要) Studies in the past provides evidence of confidence in connection with news sentiment and financial assets trend. Zhang and Skiena (2009) uses news data including the companys’ news release frequency and sentiment generated by a large-scale Natural Language Processing (NLP) news analysis system to predict or reflect on its stock return and trading volume. The study also formulate trading strategies based on the news market. Feuerriegel and Prendinger (2016) confirmed that news sentiment can explain stock price movements and construct trading strategies with news text mining.The empirical study is also based on the need to explore whether sentiment analysis has an impact and predictive power on financial asset trends. The empirical study uses several supervised machine learning classification methods, including logistic regression, random forest, gradient boosting, adaptive boosting, and support vector machine. I predict the future rise and fall of Taiwan index futures and look for the model with the best predictive power to construct the trading strategy.The empirical result shows that the addition of news sentiment variables can effectively improve the predictive power of all models. This empirical study selects the best model to construc trading strategy with the best forecast of the day and the next day. The performance of the test set can beat the buy-and-hold strategy and momentum investment strategy and the strategy predicts that the strategy of the next day`s change is better than the forecast of the day`s change. The news sentiment has the effect of delaying the emergence, and the strategy of using the gradient boosting model is outperforming. en_US dc.description.tableofcontents 目 次第一章 研究緒論 71.1 研究動機與研究目的 72.2 研究架構 8第二章 文獻回顧 92.1 人工智慧及機器學習 92.2 情緒分析應用於金融商品報酬預測 12第三章 研究方法 133.1 變數介紹 133.2 機器學習方法 163.3 模型衡量 213.4 策略績效計算與衡量 233.5 交易策略建構流程 26第四章 交易策略實證分析 284.1 模型建構與模型效果 284.2 策略績效衡量與分析 30第五章 結論與建議 405.1 結論 405.2 研究限制與未來研究建議 40參考文獻 42 zh_TW dc.source.uri (資料來源) http://thesis.lib.nccu.edu.tw/record/#G0106358006 en_US dc.subject (關鍵詞) 機器學習 zh_TW dc.subject (關鍵詞) 情緒分析 zh_TW dc.subject (關鍵詞) 新聞情緒 zh_TW dc.subject (關鍵詞) 交易策略 zh_TW dc.subject (關鍵詞) Machine learning en_US dc.subject (關鍵詞) Sentiment analysis en_US dc.subject (關鍵詞) News sentiment en_US dc.subject (關鍵詞) Trading strategy en_US dc.title (題名) 情緒分析交易策略設計 zh_TW dc.title (題名) Formulate a Trading Strategy Using Sentiment Analysis en_US dc.type (資料類型) thesis en_US dc.relation.reference (參考文獻) Antweiler, W., & Frank, M. Z. (2004). Is all that talk just noise? The information content of internet stock message boards. The journal of Finance, 59(3), 1259-1294.Bellman, R., & Lee, E. S. (1978). Functional equations in dynamic programming. Aequationes Mathematicae, 17(1), 1-18.Breiman, L., Friedman, J., Olshen, R., & Stone, C. (1984). Classification and regression trees. Wadsworth Int. Group, 37(15), 237-251.Chan, W. S. (2003). Stock price reaction to news and no-news: drift and reversal after headlines. Journal of Financial Economics, 70(2), 223-260.Cortes, C., & Vapnik, V. (1995). Support-vector networks. Machine learning, 20(3), 273-297.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.Friedman, J. H. (2001). Greedy function approximation: a gradient boosting machine. Annals of statistics, 1189-1232.Feuerriegel, S., & Prendinger, H. (2016). News-based trading strategies. Decision Support Systems, 90, 65-74.Haugeland, J. (1985). Artificial intelligence: the very idea. In: Cambridge, MA: MIT Press.Ho, T. K. (1995). Random decision forests. Paper presented at the Proceedings of 3rd international conference on document analysis and recognition.Keating, C., & Shadwick, W. F. (2002). A universal performance measure. Journal of performance measurement, 6(3), 59-84.Machinery, C. (1950). Computing machinery and intelligence-AM Turing. Mind, 59(236), 433.Malkiel, B. G., & Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. The journal of Finance, 25(2), 383-417.Shannon, C. E. (1948). A mathematical theory of communication. Bell system technical journal, 27(3), 379-423.Sharpe, W. F. (1994). The sharpe ratio. Journal of portfolio management, 21(1), 49-58.Sortino, F. A., & Price, L. N. (1994). Performance measurement in a downside risk framework. the Journal of Investing, 3(3), 59-64.Sul, H., Dennis, A. R., & Yuan, L. I. (2014). Trading on twitter: The financial information content of emotion in social media. Paper presented at the 2014 47th Hawaii International Conference on System Sciences.Tetlock, P. C., Saar‐Tsechansky, M., & Macskassy, S. (2008). More than words: Quantifying language to measure firms` fundamentals. The journal of Finance, 63(3), 1437-1467.Wilder, J. W. (1978). New concepts in technical trading systems: Trend Research.Young, T. W. (1991). Calmar ratio: A smoother tool. Futures, 20(1), 40.Zhang, W., & Skiena, S. (2009). Trading strategies to exploit news sentiment. Submitted for publication. zh_TW dc.identifier.doi (DOI) 10.6814/NCCU201900710 en_US
