Member promotion in social networks via skyline

Abstract

Research on Social Network is becoming popular in recent years because of the prevalence of online social communities, e.g., Facebook. As a result, the importance of the role which every single member plays in the social network is attracting more and more attention spontaneously. Meanwhile, the skyline operator is considered to be a useful tool to recognize important data entity, especially multi-dimensional data. The skyline operator is thus introduced into social networks to distinguish the important members from the entire community. For decision-making, people are not only interested in the result of skyline query on social networks, i.e., skyline points which represent important entities, usually known as stars, e.g., competitive companies, popular actors, famous scholars, but in many cases, people also care about the potential stars which are less important right now but will play a significant role in the near future. In this paper, we study and provide better approaches to a newly raised problem called Member Promotion in Social Networks, which aims at discovering the most potential star(s), i.e., the member(s) which will be promoted into skyline with minimum cost, in the social network for promotion. Based on the edge-weighted characteristic of the social network and the peculiarity of the promotion process, we systematically divide the scenario of the problem into two parts, i.e., unequal-weighted social networks and equal-weighted social networks. Against unequal-weighted social networks, based on some new proposed concepts, e.g., Promotion Boundary and Plan Limitation, and the theories of permutation and combination, we design a remarkably effective pruning process and propose the Skyboundary algorithm. Then, after bringing forward an interesting new concept named as Infra-skyline, we present another optimized approach, namely the Infrasky algorithm, to the member promotion problem in equal-weighted social networks. Extensive experiments on real datasets are conducted to show our algorithms can effectively find the potential members and improve the efficiency by orders of magnitude than the baseline algorithms. We also implemented a demonstration tool entitled as “StarMiner” to present the promotion process in a graphical and interactive way.