Spatial Modeling of COVID-19 Morbidity/Mortality using Machine-Learning Algorithms
Abstract
The COVID-19 pandemic has posed substantial costs on individuals and societies, both directly due to human physical and mental health and indirectly through economic and social restrictions. In this study, we have developed a geo-database of a broad range of climatic, environmental, socio-economic, and demographic variables along with pre-existing co-morbidities that can potentially contribute to the elevated COVID-19 morbidity/mortality rates. We then employed several machine learning algorithms, particularly artificial neural network topologies, to model and predict the latest status of the disease as of August 2021 in a geographic information system (GIS) framework. The accuracy of the models has been examined and validated with the comparison with ground-truth data. Our results indicated a decent prediction accuracy compared to traditional techniques. Moreover, our results suggested that when coupled with GIS, machine learning algorithms can significantly improve spatial prediction accuracy. Moreover, the most contributing variables have been identified using sensitivity analysis. The findings can provide valuable insights for public health policy makers to identify potential risk factors associated with the COVID-19 for targeted interventions.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFMGH15F0634M