Potential links by neighbor communities

The probability of two nodes to be linked is related to their similarities in the network. Based on statistical inference, a network-structure similarity index, therefore, is proposed to find the potential links. This index quantifies the effects of the node communities on these links. And an algorithm for the index is also successfully designed. The experiments on several networks with ground-truth groups and temporal attributes reveal that two nodes are likely to be connected if some of their neighbor nodes are in common communities. The results from these experiments with tested networks, several of which cover more than a million nodes, show the reliability of the index and the advantage of its algorithm.