Social Sensors in Epidemiological Networks via Graph Eigenvectors

Abstract: In this paper, we consider epidemiological networks which are used for modeling the transmission of contagious diseases through a population. Specifically, we study the so-called social sensors problem: given an epidemiological network, can we find a small set of nodes such that by monitoring disease transmission on these nodes, we can get ahead of the overall epidemic in the full population? In spite of its societal relevance, there has not been much statistical work on this problem, and we aim to provide an exposition that will hopefully stimulate interest in the research community. Furthermore, by leveraging classical results in spectral graph theory, we propose a novel method for finding social sensors, which achieves substantial improvement over existing methods in both synthetic and real-world epidemiological networks.