Boosting Tree Trading Strategy on Blue Chips in Taiwan Stock Exchange

• Main Authors: LIU, YAN-YU, 劉彥佑
• Other Authors: CHANG, CHIEN-HUNG
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/ecegc8

碩士 === 靜宜大學 === 財務工程學系 === 107 === Machine learning methods have been widely used in various industries in modern times, and they have performed well in Graphic recognition and trend prediction. In this thesis, we will use XGBoost (Extreme Gradient Boosting) as the training model algorithm, with six of the eight technical indicators , which in ADX, MACD, RSI, SAR, KD, OBV, EMV, ATR as the learning characteristics, so that the whole Models can use these characteristics to predict future trends in stock prices. Then we use the predicted results to make transactions on the current day, 5th day, 10th day, and 20th day. As a result, the combination of RSI, ADX, SAR, MACD, KD, and EMV performs better than the other groups, regardless of the average rate of return or the Sharpe ratio. Finally, it is found that the four indicators of KD, RSI, EMV and ATR are the most important for training the XGBoost model, so it will bring lost of benefit to selecting indicators .