Membership Detection for Real-world Groups Hidden in Social Network

Abstract

Real-world groups are organizations or communities existed in the real world, such as the employees of a company, the students of a school, different from the virtual communities in social networks. The members of a real-world group may also appear in the social network and form into a virtual community. However, the community detection methods are not effective to detect the real-world groups because the members may lack interaction and sensitive attributes in the social network, so that the real-world groups appear to be hidden in the social network. This paper defines three kinds of real-world group models and defines sensitive attributes and sensitive relationships of users in real-world groups. We use random walk to detect memberships for real-world groups hidden in social network with no or little edges and sensitive attributes. We evaluate our model with a Facebook dataset. The experiments show that our model has an accuracy of 95%.