Trade the Tweet: Social Media Text Mining and Sparse Matrix Factorization for Stock Market Prediction

Abstract

We re-examine the problem of using textual information to provide insights into stock market prediction. Our model differs from past models as it seeks to leverage market information contained in high-volume social media data rather than news articles and does not evaluate sentiment. Utilizing the latent space model from Wong et al. (2014), we correlate the movements of both stock prices and social media content. We extensively test this model on data spanning 2011 to 2015 on a majority of stocks listed in the S&P 500 index. Finally, we compare our model to traditional econometric models and propose a trading strategy with a Sharpe Ratio of 1.35 and an annualized return of 18%.