Sensor-Based Localization of Diffusion Sources on Networks: a Bayesian Approach

• Activity Number: 415 - Methods for Functional or Network Data
• Type: Contributed
• Date/Time: Tuesday, August 1, 2017 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Jun Li* and Juliane Manitz and Enrico Bertuzzo and Eric Kolaczyk
• Companies: Boston University and Boston University and University Cà Foscari Venice and Boston University
• Keywords: complex network ; source detection ; sensor placement ; epidemics ; KwaZulu-Natal ; cholera outbreak

Abstract

The detection of diffusion sources in complex networks based on observer data is of considerable interest. Here, we introduce a Bayesian extension of a well-established Gaussian source estimator (Pinto et al 2012), which can incorporate prior knowledge efficiently and does not require the network to have tree structure. We provide a theoretical approximation of source estimator localization probability when the observers are sparse and also propose a sensor placement strategy. We illustrate our method through a case study of the 2000 cholera outbreak in the KwaZulu-Natal province, South Africa. Because the human mobility has been recognized to be a key driver of the infectious disease outbreak (Mari et al 2011, Finger et al 2016), we combine it with a hydrological network and ecological corridors, and implement our method on the aggregate network. Our reasonable prior based on epidemic knowledge plays a significant role. This analysis has both methodological and practical importance.