Community Evolution Prediction in Dynamic Social Networks Using Community Features' Change Rates

In this paper, we investigate the prediction of community future occurring events in dynamic social networks, based on change rates of features that describe a community throughout its evolution life-cycle rather than absolute values of features. Besides, we explore the most predictive features for each event. Our experiments on DBLP and Facebook datasets, using community structural features and its influential members' features, confirm that the prediction of the next event that may occur to an evolving community using change rates of features can be achieved with a very high accuracy. We further observe that the most significant features vary according to each event prediction.