Prediction of stock price movements based on concept map information

Visualization of textual data may reveal interesting properties regarding the information conveyed in a group of documents. In this paper, we study whether the structure revealed by a visualization method can be used as inputs for improved classifiers. In particular, we study whether the locations of news items on a concept map could be used as inputs for improving the prediction of stock price movements from the news. We propose a method based on information visualization and text classification for achieving this. We apply the proposed approach to the prediction of the stock price movements of companies within the oil and natural gas sector. In a case study, we show that our proposed approach performs better than a naive approach and a bag-of-words approach.