Machine Learning Model and COVID-19 Forecasting
Machine learning algorithms have long been applied to detect risk factors and make projections for decision making. The Population Council’s HTF supports innovative machine learning projects that identify factors associated with outbreaks to understand and predict COVID-19 incidence and mortality. This research identifies societal patterns for transmission and vulnerability and can be used to develop mitigation strategies.
Geospatial Human-Centered Artificial Intelligence Lab, UC-Boulder
To learn more, visit:
Python code for predicting COVID-19 cases at the country-level in the US, Geospatial Human-Centered Artificial Intelligence Lab, University of Colorado Boulder
In the news:
Professor Morteza Karimzadesh said the data helps him and his team measure people’s movements and connectivity between locations. 9News
Predictive Power of Machine-Learning Model using Facebook Social Connectedness and Cell-phone Mobility Data
Spatiotemporal prediction of COVID-19 cases using inter- and intra-county proxies of human interactions, Nature Communications, 2021
The Population Council, in collaboration with researchers at the University of Colorado, built and tested models to predict COVID-19 cases at the county level in the United States. Researchers looked at two proxies for human interaction, the primary cause of disease transmission. Comparing Facebook friendship data against cell phone-based mobility data, the study team validated that the new models predicted, with high comparative accuracy, cases of COVID-19 in the short and long term. Accurate forecasts can help decision makers to understand and anticipate needs. Future research is now underway to forecast hospitalization cases, both for COVID-19 and influenza.