Forecasting an infectious disease: COVID-19

Details

Supervisors

Absil, Pierre-Antoine

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en mathématiques appliquées, à finalité spécialisée

Abstract

Over the past two years, many scientific papers have been published about the pandemic of COVID-19. In particular, forecasting the number of infections or the number of hospitalizations due to COVID-19 has been a topic of great interest. Indeed, those forecasts and the study of the pandemic in general have been very useful to the governments who had to take political and economical decisions to slow down the spread of the disease. Nowadays, a lot of models exist to produce those forecasts, based on the SIR model. In this thesis, we propose to further the study of one of them: the SH model. We experiment the use of the Savitzky-Golay filter to estimate one of the initial states of the model, hoping for a better accuracy in the forecasts. And, we produce confidence intervals for the forecasts, according to three sources of uncertainty of the SH model. Three different methods are developed to produce the confidence intervals while taking into account the stochastic, parameter and model uncertainty.