Temporal Dependency Between Evolution of Features and Dynamic Social Networks

Abstract

The complexity of analyzing dynamic social network data is higher than static networks because of the nature of the data itself. Nodes and links of dynamic social networks are unlabeled, not identically distributed and the effect of different features of the network evolves over time with the evolution of the network links. Changes in features of nodes initiate the changes in links and groups of the network and vice versa. The correlation between these changes is not parallel which makes things more complicated. The temporal effect increases the complexity of the evaluation by adding another vital parameter to the problem. In this paper, the effect of the temporal dependency on dynamic social network evolution was examined using a real life social media dataset. By extracting the most prominent features of the network after a specific time period with an unsupervised feature selection method, we computed the correlation between their evolution which was not uniformly distributed over the time span.