| Summary: | 碩士 === 東吳大學 === 經濟學系 === 98 === In this article,we try to use Genetic Algorithm (GA), Genetic Algorithm Back Propagation Network (GABPN) and Genetic Programming (GP) -- A.I. methods to find out the optimal strategies under the same ups-and-downs of the stock price conditions. When it tallies with some kind of conditional ups-and-downs situations in the stock price trend, we can predict its future tendency according to the similar tracks in its historical experience.
The experimental design is aimed at the daily return rate of Taiwan TAIEX INDEX(TWSE), which downs three days in succession, and the next daily return rate as the predicted value of the goal during the last ten years (2000~2010). The empirical result shows that: (1)On the predicting error criterion, GP is the best, GABPN is the next, and GA comes after that. (2)On the direction accuracy, GABPN is the best, GP is the next, and GA comes after that. These three A.I. methods in direction accuracy forecasting ups or downs of TWSE are all above 60%, the highest even achieving 80.77%, which is higher than 50%, the rate of guessing randomly, obviously. This research demonstrates that it’s the good ways that predicting the stock price after the ups-and-downs condition by the three kinds of A.I. methods.
|
|---|
