Comprehensive influence of local and global characteristics on identifying the influential nodes

Identifying the most influential nodes is one of the most promising domains in understanding and controlling propagation processes in complex network. According to the percolation theory, there is a epidemic threshold difference between the residual network and the original network after removing a node. We think that the threshold difference can represent the node's global influence, which the absence of the node can promote or suppress the epidemic outbreak. By considering threshold differences and the local property (degree centrality), we propose a comprehensive influence method (CI) to identify the influential nodes. Comparing with the susceptible-infected-recovered model, the experimental results for nine empirical networks show that the CI method which can be applied to most networks with the different structures is more accurate than the K-shell, degree, closeness, and eigenvector centralities.