一般來說,投資人通常在股票投資時參考技術指標。然而,新聞也常被他們視為投資決策時的重要資訊來源。但新聞為一種非結構化的資料,收集與分析上相當不容易。文字探勘技術可針對非結構化資料進行轉化,成為可供分析的資訊。因此,文字探勘技術將可從大量的財經新聞中挖掘出隱含的資訊。本研究提出財務文字探勘架構,來建立股價反轉時點預測模型,實驗模型中結合技術指標與財經新聞的做法不僅具預測效益外,更獲得顯著為正的可觀報酬。最後,研究結果顯示技術指標的數值資訊與財經新聞的文字資訊是有互補效應的,在反轉時點的預測上效果顯著,投資人能夠以此研究架構及預測模型來作為投資決策的參考。
In general, financial indicators are usually used to stock investment by investors. However, stock market news is also regarded as an important information sources for their investment decisions. Since news is one type of unstructured data so that it’s not easy to be gathered and analyzed for its meaning. Text mining transforms unstructured data into useful and analytical information. Therefore, text mining also helps us to reveal the implied information from an amount of financial news. This study not only proposes a financial text mining framework but also creates a model for detecting stock price’s reversal points. The proposed approach constructs the prediction model by means of combining financial indicators and relevant news. In our experiment, positive returns are proved after a series of simulation investment. Moreover, this study has demonstrated the complementary between indicators and financial news, and is able to help investors to make right decisions in stock market.