Assimilating digital social data into epidemiological forecast models

Related Staff:

Accurate forecasting of disease outbreaks relies on understanding complex human dynamics—our mobility, social interactions and behaviour in response to factors like emergencies and mass media. While many studies have attempted to incorporate so-called “media effects” into dynamical transmission models, these models have necessarily not been highly data-driven. However, the recent “big data” explosion coming from widespread internet and social media use has made information about human behaviour ubiquitous, particularly regarding our reactions to mass media. We are using digital social media data to directly quantify the effects of mass media upon disease transmission, and then assimilating this information into predictive transmission models.

More generally, we are working to develop mathematical techniques for improving disease forecasts by assimilating these emerging digital social data streams. Such techniques will provide a framework to incorporate data on social interaction into forecasts, and will contribute to the development of models of collective human dynamics which may be used for emergency response during epidemics. Incorporating human mobility, social interactions and responses into mathematical epidemiological models will improve our understanding of the complex human dynamics which drive the transmission of diseases.

Different influenza transmission models incorporating mass media effects