The Impacts of Network Structure on User Activity Level in Online Social Networks

Abstract

Understanding how users in online social networks (OSNs) get exposed to different information and how they adopt such information is essential for OSN management and for understanding information diffusion. In particular, recent studies have demonstrated the importance of studying users' information exposure in characterizing information overload and attention competition in OSNs. However, most existing work mainly focuses on users' adoption of diffused information, with few studies investigating users' information exposure. Moreover, the impacts of the underlying social network structure on users' information exposure and adoption have not been fully addressed. In this paper, we investigate the predictability of users' activity levels in both information exposure and adoption, utilizing the structural features. Specifically, we propose a comprehensive set of structural features, consisting of global features, egocentric features and community features. Based on collected abundant video diffusion traces on a popular OSN, our empirical studies demonstrate the effectiveness of our proposed features. Finally, we formulate the prediction of users' activity levels of video exposure and adoption as a multiclass classification problem. Experiments conducted on collected dataset not only validate the performance of our proposed approach, but also provide insights into the importance of different features.