Boosting financial trend prediction with twitter mood based on selective hidden Markov models

Abstract

Financial trend prediction has been a hot topic in both academia and industry. This paper proposes to exploit Twitter mood to boost financial trend prediction based on selective hidden Markov models (sHMM). First, we expand the profile of mood states (POMS) Bipolar lexicon to extract rich society moods from massive tweets. Then, we determine which mood has the most predictive power on the financial index based on Granger causality analysis (GCA). Finally, we extend sHMM to combine financial index and the selected Twitter mood to predict next-day trend. Extensive experiments show that our method not only outperforms the state-of-the-art methods, but also provides controllability to financial trend prediction.