Integrating User Relation Representations and Information Diffusion Topology Features for Information Propagation Prediction

Information propagation prediction aims to analyze the patterns of information spreading in social networks and social media, thus understand and predict the information diffusion process. Recent researches have shown that information propagation is influenced by both group and individual relations. Existing works have mainly concentrated on group relations, including social relation and dynamic structural relation. They ignore a core group relation, i.e., the co-occurrence user relation, and an important individual relation, i.e., the user-preference relation, leading to incomplete modeling of the information propagation process. To address this issue, this paper comprehensively considers both group and individual relations, and proposes an information propagation prediction model that integrates user relation representations and information diffusion topology features. For the group relation, this paper constructs a user co-occurrence graph to learn user similarity relation representations, which are then fused with information diffusion topology features to capture group relations. For the individual relation, this paper fuses user representation and influencing factors to capture internal factors on stimulating users to share information. Experimental results show that the performance of the proposed model on two public datasets is improved, and the MAP@k and hits@k evaluation indicators on the Memetracker dataset are improved by an average of 6.54% and 2.75%, respectively.