Predicting collective sentiment dynamics from time-series social media

More and more people express their opinions on social media such as Facebook and Twitter. Predictive analysis on social media time-series allows the stake-holders to leverage this immediate, accessible and vast reachable communication channel to react and proact against the public opinion. In particular, understanding and predicting the sentiment change of the public opinions will allow business and government agencies to react against negative sentiment and design strategies such as dispelling rumors and post balanced messages to revert the public opinion. In this paper, we present a strategy of building statistical models from the social media dynamics to predict collective sentiment dynamics. We model the collective sentiment change without delving into micro analysis of individual tweets or users and their corresponding low level network structures. Experiments on large-scale Twitter data show that the model can achieve above 85% accuracy on directional sentiment prediction.