FriendBurst: Ranking people who get friends fast in a short time

Number of friends (or followers) is an important factor in social network. Attracting friends (or followers) in a short time is a strong indicator of one person for becoming an influential user quickly. Existing studies mainly focus on analyzing the formation of relationship between users, however, the factors that contribute to users' friend (or follower) numbers increment are still unidentified and unquantified. Along this line, based on users' different friends (or followers) increasing speeds, firstly, we get a number of interesting observations on a microblog system (Weibo) and an academic network (Arnetminer) through analyzing their characteristics of structure and content from the diversity and density angles. Then we define attribute factors and correlation factors based on our observations. Finally we propose a partially labeled ranking factor graph model (PLR-FGM) which combines these two kinds of factors to infer a ranking list of the users' friends (or followers) increasing speed. Experimental results show that the proposed PLR-FGM model outperforms several alternative models in terms of normalized discounted cumulative gain (NDCG).