| Summary: | 碩士 === 大葉大學 === 事業經營研究所 === 91 === The present study makes use of artificial intelligence in which genetic algorithm is combined with neural networks to forecast the fluctuation direction of the stock index of China-Concept stock.
Sampling was selected between 1993 and 2002 period in which China-concept stock weighted stock indices were issued to convert to day, weed and month performance rate, with past three stock prices(YT-1, YT-2＆YT-3) forming the forecasting variables. In the process, Beta coefficient was analyzed in the China-concept stock and TAIEX. Thereafter, the genetic algorithm model in combining with neural network was trained for the sample, then pre-testing and forecasting were performed.
The reseats indicate that beta coefficient for China-concept stock and TAIEX, showed the former is better than the latter. Fluctuation direction of stock index for short-term does not reform as well as long-term. In addition, forecasting combination for Yt-1, and Yt-3 in batter than other combination in forecasting.
|
|---|
