Project R-16479

Optimizing Arbovirus Prevention and Control in the European Union: A Data-Driven Reinforcement Learning Approach (Research)

Mosquito-borne arboviruses pose a substantial threat to public health, as they are associated with a significant morbidity and economic burden. Due to climate change, such viruses are expected to cause new waves of public health emergencies in the European Union (EU). To study risk assessment and mitigation options for arboviruses with potential to invade the EU, such as Zika and Oropouche virus, an interdisciplinary approach is warranted. Firstly, a new model needs to be developed that captures the geospatial aspects and local dynamics of virus propagation. Secondly, as we are modelling transmission dynamics of a pathogen that is not yet introduced in the EU context, we can rely only on proxy data (i.e., data that is not directly related to the specific epidemiological setting at hand) to inform the model, which requires a new fitting methodology. Thirdly, as making decisions about such complex settings is hard, there is a great potential to use reinforcement learning (RL) to support the decision maker. To this end, we need to develop new multi-criteria RL algorithms, with a focus on explainability to convey the learnt policies to decision makers. Using these new methods, we will study the invasion risk of Zika and Oropouche virus in the EU and investigate mitigation and surveillance policies that cover a wide range of criteria, including: morbidity, environmental concerns, and cost-effectiveness.

01 January 2026 - 31 December 2029