Skip to Main Content
Collective synchronous behavior is a pervasive phenomenon that has attracted many researchers' interests over past decades. It can be observed in many areas easily, including biology, chemistry, physics and social society. A series of interactive processes in-between individuals trigger the formation of collective behavior. Traditional data mining methods, however, mainly concentrate on the analysis of individual behavior but ignore the potential associations. Similarly, in sociology, many well-known models based on survey sampling are not suitable for the new emerging social media platform any more, where huge amounts of data are generated by users every day. It is necessary for researchers to develop effective approaches for sampling and modeling the collective behavior on social media. In this paper, we propose an innovative model that consists of multiple hidden Markov chains. By learning a group of time-series behavior data, our model can not only predict the synchronous state of a collective, but also measure the dependency property, namely reactive factor, of each individual. Preliminary experimental result shows that CoSync model has the power to distinguish behavior patterns of different persons.