Summary: | 碩士 === 國立臺灣科技大學 === 資訊工程系 === 101 === The main purpose of this paper is to provide a prediction system for stock market and focus on Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX). This system consist three models, including HMM, GMM and SVM. In this paper, the Monte Carlo method and roulette-wheel selection are adopted to provide a more accurate output for the probability matrix of HMM and GMM. For SVM, a rolling prediction method is proposed to train model and rapidly reflect the data volatility. Finally, through a decision making process, this system could integrate the advantages of HMM and GMM based on the SVM classification.
This prediction system is not only suitable for the stock market. For the other areas with volatile data, such as temperature, air pollution, inventory forecasting, etc. This prediction system can also provide a more accurate prediction result, no matter under a stable or an instable trend.
|