On Using Cashtags to Predict Companies Stock Trends

Different theories state that future market values strongly depend on psychological and financial factors: when investors feel positive moods they invest and the value of the stock market increases; conversely, when they feel negative moods they do not invest and the value of the stock market decreases. Today, researchers are trying to exploit the data publicly available in social media and, in particular, different researches showed a connection between Twitter messages and the stock market index. In this paper, we do not focus on a generic stock market index, nor we focus on the sole sentiment analysis. Instead, our goal is to investigate whether tweet messages can be used to predict the future trend (e.g., positive, negative or neutral) of the stocks of specific companies listed in the Dow Jones stock market. In particular, we focus on companies belonging to three different economic sectors (technology, service and health-care) and we consider the trend of 5 different metrics for each stock (e.g., highest, lowest, opening price, etc.) and the trend of 13 different variables of the tweets (e.g., volume, sentiment, tweets with links, etc.). Through an evaluation that employed more than 800,000 tweets, we show that some of the proposed ad-hoc prediction methods well predict (i.e., up to 82% of success) the next day trend of the stock values of specific companies.