Event-based structural change detection in urban-scale contact network

Abstract

© 2017, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. The detection of structural changes is an important task in analyzing network evolution, especially for interactions between people, that may be driven by external events. Existing work relies on snapshot data and misses out some key functions of networks. Here, we study contact network evolution where no snapshot data are available. In spite of the challenge, this study demonstrates how contact networks can be used to predict and control infectious disease epidemics. We first model structural changes in contact networks during the 2009 influenza pandemic in Hong Kong, and then present a probabilistic framework to address it, aiming to answer when and how the underlying structure changes, utilizing multiple data sources including demographic data, and epidemic surveillance data. The efficacy and public health utility of the method are demonstrated using both synthetic and real data.