| Summary: | 碩士 === 國立中山大學 === 資訊工程學系研究所 === 105 === We can study historical stock series to predict the price trend in the future. In this thesis, we utilize 13 trend rules with three sets of parameter combinations (short, mixed and long) to generate trend signals. Then, we implement three weighted functions (unweighted, plain-weighted and EG-weighted) to combine thousands of trend signals generated by various parameters and to determine the consensus trend signal for the current day. Every two consecutive consensus trend signals forms one of four trading operations, including buy, sell or hold or no-action. We take Lee’s [16] data set for our experiment. The trading period is from 2000/1/4 to 2012/12/28 in after-hours trading. The accumulated return of the buy-and-hold method is 287.0% (average annualized return is 10.97%). Lee’s best accumulated return is 685.31% (annualized 17.18%). The best accumulated returns of our unweighted, plain-weighted and EG-weighted functions are 1154.19% (annualized 21.48%), 970.66% (annualized 20.01%) and 915.61% (annualized 19.52%), respectively.
|
|---|
