股票市場的走勢一直是許多民眾所關注的話題,而對於股票市場上股價因素的漲跌,有許多投資人希望對於未來股票價格進行準確預測,但股價受到很多因素;包含人為因素、政治因素、總體經濟環境因素、消息面或是其他未知因素的影響,使得許多相關研究的模型準確率無法令人滿意。 Data mining技術應用於預測的延革已經很久,而過去研究使用的方法包含基因演算法(Genetic Algorithms)、類神經網路(Artificial Neural Network,ANN)、決策樹演算法(Decision Tree Algorithms)、支援向量機(Support Vector Machine)等等均有學者應用於建立股價預測模型,近年來決策樹演算法成為在股價預測的議題中受歡迎的一種,研究顯示也有不錯的效果。 本研究應用決策樹演算法於台灣集中及櫃檯市場之類股股價的預測,研究目標為台灣50類股,研究期間為西元2002年5月2日起至2010年3月31日止。 實驗結果發現,以決策樹演算法結合個股技術指標及籌碼指標變數建立的股價漲跌預測模型之投資報酬率,可與在同期間之投信基金報酬率之排名獲得不錯的成效,同時領先相同期間之大盤指數漲幅。
The trend of stock market is a subject to be concerned by most people. Many investors expect to predict the stock price, which is affected by lot of factors, such as human interference, political issue, macro-economic or other unknown factors. Therefore the accuracy of some related models cannot be satisfied by people who concern. Some methods applied to establish the stock price forecasting model by scholars, including Genetic Algorithms(GA), Artificial Neural Network(ANN), decision tree algorithms, SVM(Support Vector Machine) and so on. However, data mining technology has been used to forecast for a long time. Recently, decision tree algorithms become a popular one in the study of the stock price prediction and the study has shown good results. This research applies decision tree algorithm to forecast the stock price of listed and OTC market in Taiwan. The target of research is to study TSEC Taiwan 50 sectors during the period from 2nd May 2002 to 31st March 2010. Experimental results show that the ROI (return on investment) of the stock price index prediction model established by combining the decision tree algorithm with the technical and counter indexes variables of stock can get a good result comparing to the ranking of the return on ROI of Trust Fund in the same period; and also ahead to the increase of market index during the same time period.