Abstract
In this paper, we present algorithms to find near-optimal sets of epidemic spreaders in complex networks. We extend the notion of local-centrality, a centrality measure previously shown to correspond with a node's ability to spread an epidemic, to sets of nodes by introducing combinatorial local centrality. Though we prove that finding a set of nodes that maximizes this new measure is NP-hard, good approximations are available. We show that a strictly greedy approach obtains the best approximation ratio unless P = NP and then formulate a modified version of this approach that leverages qualities of the network to achieve a faster runtime while maintaining this theoretical guarantee. We perform an experimental evaluation on samples from several different network structures which demonstrate that our algorithm maximizes combinatorial local centrality and consistently chooses the most effective set of nodes to spread infection under the SIR model, relative to selecting the top nodes using many common centrality measures. We also demonstrate that the optimized algorithm we develop scales effectively.
Highlights
In this paper we look to find optimal sets of individuals in a complex network to initiate an epidemic
Addressing such a problem will have clear implication in seeding a social network to ensure a given phenomenon diffuses optimally and may provide insight into mitigation strategies against an infection initiated by a group of individuals
We explored the problem of identifying a set of nodes that will cause an epidemic to spread under the SIR model of [7]
Summary
In this paper we look to find optimal sets of individuals in a complex network to initiate an epidemic. Addressing such a problem will have clear implication in seeding a social network to ensure a given phenomenon diffuses optimally and may provide insight into mitigation strategies against an infection initiated by a group of individuals. The susceptible-infected-recovered (SIR) model [3] is one of the most well-studied models of epidemic disease spread in a population. There has been much interest in studying this model on populations structured as a network [4,5,6]
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have