News Sensitive Stock Trend Prediction

Stock market prediction with data mining techniques is one of the most important issues to be investigated. In this paper, we present a system that predicts the changes of stock trend by analyzing the influence of non-quantifiableinformation (news articles). In particular, we investigate the immediate impact of news articles on the time series based on the Efficient Markets Hypothesis. Several data mining and text mining techniques are used in a novel way. A new statistical based piecewise segmentation algorithm is proposed to identify trends on the time series. The segmented trends are clustered into two categories, Rise and Drop, according to the slope of trends and the coefficient of determination. We propose an algorithm, which is called guided clustering, to filter news articles with the help of the clusters that we have obtained from trends. We also propose a new differentiated weighting scheme that assigns higher weights to the features if they occur in the Rise (Drop) news-article cluster but do not occur in its opposite Drop (Rise).