Infectious disease forecasting plays a critical role in outbreak response, guiding public health interventions and resource allocation. However, many existing forecasting models rely on oversimplified assumptions, such as constant reproduction numbers (Rt), which limit their utility for long-term predictions. Furthermore, while ensemble methods have proven effective in other fields, their systematic application to epidemiological forecasting, including learning from past performance, remains underexplored. Evaluation approaches often lack focus on the specific aspects of predictive performance that are most relevant to decision-making.

This project seeks to improve infectious disease forecasting by addressing three interconnected areas. First, it will develop models that move beyond stationary assumptions, enabling forecasts to better account for dynamic changes in disease transmission. Second, it will systematically investigate ensemble methods, including advanced weighting schemes such as those implemented in tools like the qraensemble and stackr R packages, to combine predictions from multiple models more effectively. Third, it will design evaluation frameworks tailored to decision-making, focusing on practical metrics such as outbreak peak timing, magnitude, and critical thresholds. The research will be applied to case studies including COVID-19, mpox, influenza and other infectious diseases, ensuring the methods are rigorously tested and broadly applicable, and aim to improve the utility of infectious disease forecasts in future epidemics.

The project would suit anyone keen to get insights into the application of quantitative techniques in public health contexts, specifically advanced analytics applied in epidemiological contexts, as well as broader understanding of forecasting and forecast evaluation. It will be jointly supervised by Sebastian Funk (LSHTM), Anne Cori (Imperial) and Edwin van Leeuwen (UKHSA). Students are expected to have a postgraduate degree, ideally in a quantitative subject (e.g. Biostatistics, Bioinformatics, Mathematics, Statistics, Computer Science or Physics) or a related discipline (e.g. Epidemiology or Biology) with a strong quantitative element either awarded or imminent or equivalent training. At least some coding experience, ideally in R, is also required.