Approximate Bayesian Computation for Data-Driven Epidemiological Models

University essay from Uppsala universitet/Institutionen för informationsteknologi

Author: Christoph Nötzli; [2023]

Keywords: ;

Abstract: Epidemiological models can help us to understand the spread of pathogens in a population. Fitting these mathematical models to epidemiological data can be a difficult task due to uncertain or missing data. Simulation-based approaches, such as approximate Bayesian computation with sequential Monte Carlo sampling, are a popular way to deal with parameter inference problems, when the likelihood is untraceable or infeasible to compute. Approximate Bayesian computation is flexible approach and can be optimized to specific problems. We investigate the possibility to optimize the perturbation kernels, the summary statistics and the parameters within the simulations for the epidemiological use case. The optimized algorithm shows promising results with the true value mostly staying within the 95% confidence interval of the parameter estimates.

  AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)