MPPR: Multi Perspective Page Rank for User Influence Estimation

With the rapid development of social networks, it is important to identify users with high influence. In content curation social networks (CCSNs), there are two kinds of user relations. The one is the explicit user relations from user's following behavior. And the other is the content based user relations from user's repin behavior. Based on these observation, we propose multi perspective page rank (MPPR) to estimate user influence. The proposed algorithm integrates both user relations to calculate influence scores of the users automatically. User influences are computed based on the transition matrix of following and repin relations. When the iteration is convergent, every user will get a fixed influence value. Experiments on the dataset containing 11990 users, 920610 following relations and 39321016 repin relations show that the proposed algorithm outperforms the typical PageRank algorithm.