Analyzing and predicting news popularity on Twitter

Twitter is not only a social network, but also an increasingly important news media. In Twitter, retweeting is the most important information propagation mechanism, and supernodes (news medias) that have many followers are the most important information sources. Therefore, it is important to understand the news retweet propagation from supernodes and predict news popularity quickly at the very first few seconds upon publishing. Such understanding and prediction will benefit many applications such as social media management, advertisement and interaction optimization between news medias and followers. In this paper, we identify the characteristics of news propagation from supernodes from the trace data we crawled from Twitter. Based on the characteristics, we build a news popularity prediction model that can predict the final number of retweets of a news tweet very quickly. Through trace-driven experiments, we then validate our prediction model by comparing our predicted popularity and real popularity, and show its superior performance in comparison with the regression prediction model. From the study, we found that the average interaction frequency between the retweeters and the news source is correlated with news popularity. Also, the negative sentiment of news has some correlations with retweet popularity while the positive sentiment of news does not have such obvious correlation. Published by Elsevier Ltd.