Stock prediction using multiple time series of stock prices and news articles

In the stock market, stock prices of multiple companies interact with each other. For instance, a stock price movement of a company triggers that of another one. Therefore, investors are interested in inter-relationship of multiple companies whose stock prices interact with each other. In recent years, a number of studies are conducted to predict stock price movements in the area of artificial intelligence. Most of them focus on stock prediction but not on explaining the reason why they succeed in stock prediction. In this study, we propose a method to find out a rule that predicts the stock price movement of a target company. We use rate of change of multiple companies´ stock prices and a newspaper article about a company in the rule. We explain the reason why the rule succeeds in its prediction by analyzing inter-relationship of these companies and the target company by the use of newspaper articles and stock prices related to Companies Co-occurrence Map. This method is applied to the first section of the Tokyo Stock Exchange and encouraging results are obtained.