Dynamic prediction of forthcoming trends in stock prices from news articles
Wei Fan, Toyohide Watanabe · 2012
In this paper, given a news article referring to one company, we decide whether it is a piece of good news that is followed by a moving-up trend in the company's stock market or a piece of bad news reversely. Additionally, we predict how will the fluctuation of stock price be influenced by the news article. The existing research work did not support flexible identification of the trends in stock price series, or take account of the case that temporal consecutive news articles may influence the stock market sensitively. In our proposed methods, we realize a more flexible and accurate investigation of correlation between news articles and stock prices. Experiments of our proposed methods yield high accuracy of prediction. The proposed mechanism for dynamically choosing sliding window to identify trends is also proven to be effective.