近年來，有許多的研究希望透過市場上現有的消息，例如公司的財務報表以及新聞報導，來進行對股價的預測及走向判斷。根據Fama提出的效率市場假說[12]，這些公開的訊息會反映在股價的變化上，因此，如何從這些訊息中擷取出有效的內容以判別股價的走向是此類研究的重點。然而在這一方面，過往的研究大多以詞袋(Bag of words)來進行模型的建立，再進一步則是使用複合詞(complex words)如N-gram、名詞短語(Noun phrase)等等。較少有研究進一步的在文中去搜尋與股價有確切關聯的內容。在本研究中，我們使用了不同的文本探勘工具去找尋與特定公司更具有關聯性的內容，並分析這些內容與其股價的關係。我們希望透過應用近年來不斷成熟的文本探勘技術，針對新聞中單詞的詞性、句子的結構以及情感分數進行更有效的特徵擷取，以增進預測模型的精準度。;In recent years, there are many studies try to predict the direction of stock price with available message on the market, such as financial statements and financial news. According to Fama′s efficient market hypothesis[12], these public information will be reflected in the change of stock price. Therefore, how to retrieve the effective message from news to determine the stock price trend is the significant point of such research. However, in this aspect, past studies mostly established prediction model with bag of words, still further was the use of complex word such as n-gram, noun phrase, etc, few studies have further to search the text content associated with the stock price in the news. In this study, we used some text mining tools to find the more relevant content of specific company and analyze the relationship between these content and the company’s stock price. We hope to get effective features through applied the more mature text mining technology for part of speech of words, sentence structure and sentiment analysis, we can enhance the accuracy of the prediction model.
