Friend Recommendation Based on the Similarity of Micro-blog User Model

Online social network has obtained a significant increase in recent years. Making friends is an ordinary way of establishing social relationships with others in online social network. Therefore friend recommendation is becoming a very important aspect and attracting extensive attention in visual communities and social network. Among various online social networks, micro-blog has become increasingly popular. The rapid growth of micro-blog data provides a rich resource for social community mining. In this paper, we present a friend recommendation approach based on the similarity of micro-blog user model. The proposed approach constructs the user model by considering four aspects: user profile, the content information that user posted, the link relationship and the interaction relationship between users. The interaction information such as users' comments and forwards are considered while calculate the link strength between users. The micro-blog content is divided into different topics which are obtained from an existing topic hierarchy. After we obtained the micro-blog user model, we calculate the interaction based similarity and content based similarity separately. After that, we mix the two kind of similarity together to calculate the similarity of micro-blog user model. At last, we recommend friend to users by this similarity. Experiments show that the friend recommendation approach based on the similarity of micro-blog user model has a higher precision than traditional approaches.