Abstract:  Existing methods are almost based on the single factor in characterizing the relationship within the cluster which cannot truly reflect the actual Web cluster’s structure. Aiming at the problem this paper put forward a supernetwork model to characterize the implicit relationship of Web entities corresponding to their interests. Recombining with dominant social relationships the paper constructed a more realistic multidimensional relationship model to characterize the Web cluster relations. Algorithm and experiments were given to illustrate the model’s effectiveness. The model not only characterizes a more realistic and precise network structures, but also allows detecting network hotspot information which leads a better understanding of public opinion to pave the way to effective analysis on network public opinion.

Key words: social network analysis, Web cluster, supernetwork model