Affiliations: [a]
School of management, Shanghai University, Shanghai, PR China | [b]
Department of automation, East China University of Science and Technology, Shanghai, PR China | [c] Univ Rennes, Inria, CNRS, IRISA, France
Correspondence:
[*]
Corresponding author. Shugang Li, School of management, Shanghai University, Shanghai, 200444, PR China. E-mail: [email protected].
Abstract: Selecting thepotential brand spokesperson on social network (SN), who has the huge growth potential and will own the largest number of fans in the future, can providehigher returns at lessrisks. In this study, smart link prediction algorithm (SLPA) is proposed to predict the evolvements of SNs and the brand spokesperson with the great future potential is selected based on the prediction results of SN evolution. In SLPA, mean roughness classification uniformity (MRCU) is developed to select the high efficient link prediction algorithm (LPA) for node pairs to be predicted from the local similarity based and quasi-local similarity based LPAs. MRCU uses the rough set theory and granular computing to describe the similarity of LPAs, consequently the LPA selected with MRCU can share as much similarity as possible with the other base LPAs. Furthermore, SLPA adopts the branch and bound method to smartly cluster node pairs by adaptively selecting optimal LPA for given node pairs and excluding the ones with the least possibility of linking, and consequently the most reliable results of node pairs with the highest linkable possibility are acquired. The experimental results on three SN datasets confirm the validity of SLPA in selecting band spokesperson.
Keywords: Brand spokesperson, rough set, link prediction, branch and bound
