Information Cascades over Diffusion-Restricted Social Network: A Data-Driven Analysis

Abstract

In social networks, as popular content reshares from user to user, information flows in the form of cascades. Modeling such information diffusion in social networks is of great importance and has attracted tremendous amounts of research interest. While existing work extensively studies open social networks as exemplified by Twitter and Facebook, information cascade in diffusion-restricted social networks remains much less understood. To tackle the problem above, in this paper we empirically study the embedded webpages (a.k.a. H5 webpages) spreading in WeChat Moments (WM) — a diffusion-restricted social network based on the largest Chinese social platform WeChat. WM differs from open social networks as it pays more attention to users' privacy (e.g., users' information can only be seen by their friends). In particular, we systematically analyze rich information cascades of advertising and political webpages in WM from a diffusion perspective, and further predict the growth of information cascades. We demonstrate that it is feasible to predict the scale of cascades at an early stage, and there exists a strong correlation between the cascade scale and cascade structure. This result suggests that a better understanding of information diffusion in WM is very critical, which can be further exploited to many applications such as viral marketing advertising and rumor detection.