Weighted kshell degree neighborhood: a new method for identifying the influential spreaders from a variety of complex network connectivity structures

Abstract Due to the fast and worldwide growth of the social network, it has become a potent platform for broadcasting any information. Through the network, people can easily reach to a mass, can easily propagate a piece of information within a short time. Considering the advantages, especially to accelerate the information spreading or controlling the spreading, the organizations want to exploit the social network to its best. However, as we know, the network is formed by connecting one node (i.e., user) to another node, and it is not that all the nodes will be effective equally in spreading. Because it depends on many factors and one of them is their topological position in the network. Automatically finding the effective nodes (the influential spreaders) from a network is a real challenge. In the literature, kshell decomposition and degree centrality are the two popular measures for identifying the influential spreaders from a network. Moreover, it is more challenging in identifying the influential spreaders when network connectivity structure varies from network to network. It has been found that the kshell decomposition method works better in the complete global network connectivity structures and neighbors’ degree method in the incomplete global network connectivity structures. But the degree of completeness of the network connectivity structures also vary. Under this circumstance, only the kshell method or only the neighbors’ degree method will not be able to obtain the best influential spreaders. To overcome this problem, this article proposes an indexing method weighted kshell degree neighborhood which is a composition of kshell and degree through tunable parameters. We have evaluated the effectiveness of the proposed method using different real networks and the Susceptible-Infected-Recovered (SIR) spreading epidemic model. The results show that the proposed method can significantly obtain the best spreading dynamics from different varieties of network connectivity structures and outperforms the other existing indexing methods.