Spanning graph for maximizing the influence spread in Social Networks

Abstract

In this paper, we consider the influence maximization problem in social networks. The aim is to find a subset of k -- nodes called seeds, which maximizes the influence spread. We propose a new approach based on the Independent Cascade Model (ICM) which extracts an acyclic spanning graph from the social network. The extraction method used to build the acyclic spanning graph is based on the existing centrality measures to determine the firsts nodes. We implement two extraction algorithms: SCG-algorithm for a connected graph and SDG- algorithm for a digraph. Both proposed algorithms are effective and their complexity is O(nm). So we use the same centrality measures to determine the seeds in the extracted graph. To show the pertinence of our approach, the results showed that the seeds given by the acyclic spanning graph give better results than the seeds given by the initial graph. This seeds will be determined by using the same heuristic like degree heuristic, degree discount heuristic, degree diffusion heuristic. The performances of this approach are very perceptible through the simulation carried out by the R software and the igraph package.