Interdisciplinary study on popularity prediction of social classified hot online events in China

We offer an interdisciplinary study of computer science and social science, analyzing behavior surrounding three types of online events: political events, social events, and non-public events. Based on the intrinsic characteristics of the three event types, this paper creates an effective method to predict such events. We continuously followed and recorded data every 10 min for 10 months from September 14, 2012 to July 11, 2013, and collected over 14 million “hot” posts from Sina Weibo, the largest microblogging provider in China. After removing spammers and noises, we developed a database of 4180 hot online events and 7,761,395 threads. We found that people's online behavior regarding event types varies in terms of follow-up statistics and the predictability of events. The Chinese are, typically, quite concerned with social affairs that relate most closely to their personal interests and preferences. People tend to cluster around political events more often than social events and non-public events. This is demonstrated by an algorithm embedded with a clustering growth pattern of events, which predicts the popularity of online political events above others. The statistical findings are justified by Habermas' public sphere theory and the theory of vertical/horizontal collectivism/individualism. This research provides an interesting piece of computational social science work to assist in the analysis of incentives concerning China's collective events.