Abstract: Recently the structural features of networks are widely used to the link prediction problem. Based on the information-theoretic model, we propose a more general information-theoretic model by encoding various network structural information. Specifically, for the scale-free networks, a set of Neighbor Set Information (NSI) based indices by suppressing the contribution of high-degree neighbors are proposed. Secondly, to incorporate the community information, this paper further presents a set of NSI based indices in which the prior probability of a node pair being connected is encodes the community information of networks. The experimental results on a series of real networks show that our methods outperform other classical link prediction indices.

Key words: link prediction, scale-free, network community, information entropy