Ranking Node Importance in Large-Scale Complex Network: From a Perspective of Local Abnormal Links

With the advent of big data era, the design of ranking algorithms to enhance network survivability and structural stability on large-scale networks has gained increasing attention in the recent years. In this paper, we emphasize the role of abnormal links in evaluating the importance of nodes. The abnormal links refer to those are deduced by link prediction method that the existence probability is very low yet really exist in the network, and the similarity between nodes can be used to characterize the credibility or importance of the edge that connecting the two nodes. For a pair of nodes where no link is observed in the network, the similarity between the two can be used to predict the possibility of generating link. While for node pairs that exist links, similarity index can be used to evaluate the credibility or importance of the link in the network. By taking into account the degree value and the irrationality of the existence of nodes' adjacent links, we propose a simple yet effective metric for ranking node structural importance. We evaluate the effectiveness of the proposed method on six realworld networks by static attacks and dynamic attacks, by comparing the decline rate of the relative size of the giant component of all networks, we show that the proposed method is better than other local metrics.