投資理財於股票市場為滿吸引人的途徑，因為影響股價的因素眾多便有著風險的存在。然而資訊科技不斷的發展，許多學者嘗試以各種理論、方法分析及解釋股票市場的趨勢及買賣之時機，以人工智慧方法最為引人注目。 本研究嘗試以運用支援向量機(Support Vector Machines)於台灣股票市場股價波動之預測。以公開之九家股票上市公司為樣本，用主成份分析及逐步迴歸之方法來作為技術指標的篩選，透過支援向量機來建構預測模組後再行預測出股票價格的漲跌，並進一步執行模擬投資。 研究結果顯示在某些情況下台灣股票市場是可以被預測的。另外，本研究在於模擬投資部份其結果呈現出比中華郵政定期存款利率高且呈現20%以上的報酬率。 To make money, an appealing method is to invest companies’ stocks. However, it is risky as there are many factors that effect prices of the stocks. Many theories and methods are proposed for predicting the trend of stock markets. It is hoped that with the help of modern Information Technologies, especially AI, correct buying and selling can be achieved. Our research used Support Vector Machines for predicting Taiwan stock price fluctuation. There are nine companies sampled in the study. Principal Component Analysis (PCA) and Stepwise Regression are used to choose technical indicators. Support Vector machines are applied to construct predictive models for the stock price fluctuation and investment simulation. Results reveal that Taiwan stock prices can be forecast with certain accuracy in some cases. In a trading simulation, our method produced more than 20% reward, which is better than the fixed deposit of China Postal Remittances and Savings Bank.
