A Content-Based Approach for User Profile Modeling and Matching on Social Networks

摘要

The development of social networks gives billions of users the convenience and the ability to quickly connect and interact with others for raising opinions, sharing news, photos, etc. On the road for building tools to extend friend circles as large as possible, one of the most important functions of a social network is the recommendation which proposes a group of people having some common characteristics or relations. A majority of social networks have friend suggestion function based on mutual friends. However, this suggestion mechanism does not care much about the actual interests of the users hidden in his comments, posts or activities. This paper aims to propose a profile modeling and matching approach based on Latent Dirichlet Allocation (LDA) and pretopological-based multi-criteria aggregation to explore topics that exist in user posts on a social network. We explored interesting points of pretopology concepts - a mathematical tool - and applied them for better solving the raised problem. This approach allows us to find out users who have similar interests and also other information involving user profiles.