Evolutionary Spatiotemporal Community Discovery in Dynamic Weighted Networks

Abstract

Detecting evolving communities in dynamic weighted networks are significant for understanding the evolutionary patterns of complex networks. However, it is difficult and challenging for traditional approaches to extract evolving communities with notable significance from dense and large dynamic complex networks, because most of communities are still so dense and large that we could not observe directly the detailed evolving sub-structures. In this paper, a novel approach is proposed to extract overlapping evolutionary spatiotemporal communities in large, dense and dynamic weighted networks. Evolutionary spatiotemporal communities can not only show the evolutionary of nodes and edges in a certain period clearly, but also contain weight vectors with similar evolving trend. Experiments on the global trading network show that the proposed approach can discover more sophisticated evolving patterns and properties which hide in those seemingly stable community structures.