Inferring Information Propagation over Online Social Networks: Edge Asymmetry and Flow Tendency

Abstract

Inferring the underlying information propagation over online social networks is important because it leads to new insights and enables forecasting, as well as influencing information propagation. In this paper, we analyze propagation processes in online social networks, when only limited details of propagation are available. We use data crawled from Chinese largest recommendation social network -- Douban Network, and study how the information propagates through the user relationship network of Douban. By using the users' follow relationship information, and time sequence of registering for participation in events, we build the potential propagation paths of events. After analyzing the propagation processes of 30,778 events in which about 1.47 million users are involved, we observe the statistical characteristics of propagation paths of those events, including the different types of participants and size distribution of connected participants. Further, we find that information propagation between node pairs are asymmetric. Moreover, based on the asymmetric property between node pairs, we propose a concept -- Information Potential Energy, that describes the capability that nodes disseminate information over a network. Finally, we propose a Flow Shell (FS) model that can efficiently and correctly calculate the nodes' Information Potential Energy, and validate it.