Personalized Recommendation via Relevance Propagation on Social Tagging Graph

This paper presents a novel random walk based relevance propagation model for personalized recommendation in social tagging systems. In the model, the tags are used to express the profiles of both users and resources, and then candidates of resources are recommended to the users based on the profile relevance between them. In particular, how the users to find the resources of interest is modeled as a random walk by which the relevance spreads in User-Resource-Tag relation graph. Experimental results on two real datasets collected from social media systems show the merits of the proposed approach.