Stock Market Prediction using Text-based Machine Learning

Predicting stock market price movements can be a difficult task for traditional algorithms as random events can vastly change a stock’s value. The goal of this research project is to design machine learning algorithms to predict these changes based upon communal discussion. The discussions that are being analyzed will be from forum posts of users that have varying levels of involvement with the company of focus. The posts themselves should contain information related to the current events, problems, community sentiment and other factors that would influence buyers and sellers. The proposed algorithm to make these predictions with a recurrent neural network (RNN) that will be able to analyze patterns in word use and order, placing reactions to forum posts into a category based upon expected price movement over various lengths of time. These methods show promise in predicting performance over many time frames.