Predicting the near future of infectious disease outbreaks is crucial to inform the policy response. During
the COVID-19 pandemic an enormous variety of models incorporating varying levels of detail on infection
mechanisms and integrating different amounts and types of epidemiological data were used in order to
provide infectious disease forecasts. To date it is not clear how best to combine such mechanisms with the
data available for maximising utility for decision making.

Nowcasts and short-term forecasts were a key component to the modelling response to the COVID-19
pandemic and are likely to be in high demand during future incidents. Advancing the methodology for
better situational awareness, as well as a better understanding of the role that different data streams can
play and development of tools implementing such methods would all improve future pandemic
preparedness in UKHSA.

Among the questions that could be addressed with this PhD are:
• How are state-of-the-art methods of machine learning best applied to spatially resolved infectious disease data in order to make granular predictions of the near future?
• Can mechanisms of transmission be learned from available data streams and beused to inform forecast and scenario models?
• What is the value of different data streams in informing the learning of mechanisms of transmission that inform forecasts?

The project will investigate the use of machine learning methods applied to infectious disease case counts, informed by other available data streams such as behavioural data, environmental surveillance data or individual-level measurements. They will be implemented in R, Julia, python or a similar programming language.