利用XGBOOST預測臺灣50ETF成分股股價漲跌

Abstract

本文針對臺灣50成分股中的49家公司，使用其2011/1/1至2019/12/31之盤後日資料，如：開盤價、收盤價、最高價、最低價、成交量和簡單移動平均(SMA)、指數加權移動平均(EWMA)、相對強弱指標(RSI)、隨機指標(KD)、威廉指標(Williams %R)等技術指標，作為模型輸入變數，搭配XGBoost演算法預測個股2017/1/1至2019/12/31之每日的隔三個交易日開盤價漲跌。最後再利用預測結果製作交易策略比較與買進持有策略的獲利差異 從實證結果來看，使用XGBoost搭配盤後日資料與技術指標，無法有效預測隔三個交易日的開盤價漲跌。臺灣50成分股股票市場應符合Fama(1970)提出的弱式市場效率性。

In this paper, we want to predict to the direction of stock price of FTSE Taiwan 50 Index Constituent by using technical analysis indicators and Extreme Gradient Boosting(XGBoost). We conduct an empirical analysis, investigating the stocks of FTSE TWSE Taiwan 50 Index from 2011 to 2019.In our analysis, we use daily stock close price, open price, high price, low price and volume to calculate common technical analysis indicators such as SMA, EWMA, RSI, KD and Williams %R. Then we take those indicators as inputs of XGBoost to predict the direction of stock price. At last, we develop a trading strategy to calculate the profit according to our prediction result. In our empirical result, we find that our prediction of the direction of the stock price is not precise enough to earn excess return, so the stock market of FTSE Taiwan 50 Index Constituent should be the weak form of efficient market.