Summary: | 碩士 === 國立臺灣科技大學 === 資訊工程系 === 101 === The financial time series include high-frequency, non-stationary, deterministically chaotic and contains a lot of inherently noise. Simply use only the original stock price data failed to provide satisfactory prediction of performance. Therefore, this paper using Grey Relational Analysis (GRA) and Discrete Wavelet Frame Transform (DWFT) for feature extraction, extract hidden information, to improve forecast accuracy. But different feature set of prediction models for their predictive capacity may vary. Therefore, how to select the feature set is very important. This paper using Genetic Algorithms (GA) to select the better set, last using Support Vector Regression (SVR) for training, the establishment of prediction model to predict Taiwan weighted stock index (TAIEX) for increasing the forecasting accuracy. It is evident that the proposed approach gets the better result performance than that of the other methods.
|