The Structure and Evolution of Large Cascades in Online Social Networks

Abstract

The emergence of online social network services allows user to share photo, video or other content with their social friends. The content is transmitted from person to person, and a diffusion cascade form. Many recent works have discovered that the vast majority of cascades are small and only a tiny fraction of content can spread widely. In this paper, we focus on the structure of these rare but large cascades in online social networks. We introduce the concept of combined graph which not only contains the diffusion links but also includes relevant friendship edges. We find that the characteristics of combined graph provide a deep understanding of how information flows and reaches a large population on social network. We investigate over 45000 large cascades whose sizes range from thousands to hundreds of thousands. We show the temporal dynamics of cascade tree and combined graph, and find that the combined graph is sparse, less clustering and lack of a dense core. In addition, we analyze the phenomenon from a microscopic perspective. Finally, we examine the correlations between structural properties and summarize four structural patterns.