對個體投資者的評論進行文本挖掘並預測股價漲跌

Abstract

長期以來，中國大陸股票交易市場中的 A 股市場，全體個體投資者即所謂的散戶,持股市值超過市場縂市值的55%。雖這個比例在逐年下降，但仍無法忽視個體投資者對市場的影響。大多數分析師和投資者都會選擇對專業金融機構給出的研究報告進行分析，很少會關注個體投資者的意見和態度。所以在本論文中，我希望藉助文本挖掘（text mining），分析個體投資者在財經網站對各股的貼文和評論，提取個體投資者對個股的態度是看漲還是看跌，以此觀察個體投資者的總體看法是否符合股票的走勢。實驗結果表明，通過文本材料分析出個體投資者是看漲或看跌的態度是困難的，而且標記好的文本材料也很稀缺，大大地影響了分類的準確性。因此限制了個體投資者的總體看法和股票走勢是否契合的分析。在本論文的結果來看，利用個體投資者的看法預測股票走勢效果並不好。

For a long time, more than 55% of the overall capitalization of A-shares in Mainland China is owned by individual investors. Although this ratio is declining yearly, the impact of individual investors on the stock market can not be ignored. Most analysts and investors analyze research reports from professional financial institutions, and little attention has been paid to the opinions and sentiments of individual investors. To close that gap, this thesis uses text mining techniques to analyze individual investors’ posts and comments on financial websites to categorize whether individual investors are bullish or bearish on specific stocks. In a way, we can find out whether they track the trends of specific stocks well. Results show that it is hard to identify the inclinations of individual investors towards a stock. That the data set with tags is hard to obtain makes classification even more difficult. Therefore, the research about whether individual investors can track the trend of a stock is limited. Our experimental results suggest that it may not be a good idea to make stock prediction through individual investors’ posts.