Data mining techniques to identify the direction of Taiwan Stock Index Futures day trading

• Main Authors: Shih Hsiao Cheng, 施孝承
• Other Authors: Ming-Chin Hung
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/62576979033399310486

碩士 === 東吳大學 === 財務工程與精算數學系 === 99 === This study focused on whether the price movement of TAIEX Futures in every five-minute interval of the first 105 minutes impacts on the price moving direction during the rest of trading hours. The main research tool for this study is the decision tree model in data mining technology. There are two steps of analysis procedure; first, discover meaningful rules in historical data with decision tree model. In this step, this experiment proved that the rules excavated out by decision tree model can identify efficiently “long futures on the relatively favorable direction of the trade” and “short futures on the relatively favorable direction of the trade”. The average of the accuracy rate in five years is up to 60 percent. Second, to verify the 4 scenarios of market and trading advices from cross-matching the two rules concluded from first step and decision tree model. The result showed that “trading recommendations” can’t identify efficiently the scenarios of the price movements in Taiwan futures market after 10 o’clock. However, if explain the result in the opposite aspect, the average of confirmed fault rate is less than 40 percent. Here the confirmed fault rate means proposing the advice in opposite trading direction. Therefore, the trading advice model is still valuable and which can ensure positive returns with stop-loss mechanism.