Department of Computer Science
Inferring metapopulation based disease transmission networks
To investigate how an infectious disease spreads, it is most desirable to discover the underlying disease transmission networks based on surveillance data. Existing studies have provided some methods for inferring information diffusion networks, where nodes correspond to individual persons. However, in the case of disease transmission, to effectively develop intervention strategies, it would be more realistic and reasonable for policy makers to study the diffusion patterns at the metapopulation level, that is, to consider disease transmission networks where nodes represent subpopulations, and links indicate their interrelationships. Such networks are useful to: (i) investigate hidden factors that influence epidemic dynamics, (ii) reveal possible sources of epidemic outbreaks, and (iii) practically develop and improve strategies for disease control. Therefore, based on such a real-world motivation, we aim to address the problem of inferring disease transmission networks at the metapopulation level. Specifically, we propose an inference method called NetEpi (Network Epidemic), and evaluate the method by utilizing synthetic and real-world datasets. The experiments show that NetEpi can recover most of the ground-truth disease transmission networks based only on the surveillance data. Moreover, it can help detect and interpret patterns and transmission pathways from the real-world data. © 2014 Springer International Publishing.
Bayesian learning, disease transmission networks, metapopulation, Network inference, partial correlation networks
Source Publication Title
Advances in Knowledge Discovery and Data Mining: 18th Pacific-Asia Conference, PAKDD 2014, Tainan, Taiwan, May 13-16, 2014. Proceedings, Part II
Springer International Publishing
Link to Publisher's Edition
Yang, Xiaofei, Jiming Liu, William Kwok Wai Cheung, and Xiao-Nong Zhou. "Inferring metapopulation based disease transmission networks." Advances in Knowledge Discovery and Data Mining: 18th Pacific-Asia Conference, PAKDD 2014, Tainan, Taiwan, May 13-16, 2014. Proceedings, Part II (2014): 385-399.