In collaboration with the Harvard Global Health Institute, Google today released the COVID-19 Public Projections, a set of models that supply forecasts of COVID-19 cases, deaths, ICU usage, ventilator accessibility, and other metrics over the next 2 week for U.S. counties and states. The models are trained on public data such as those from Johns Hopkins University, Descartes Labs, and the United States Census Bureau, and Google says they’ll continue to be upgraded with guidance from its collaborators at Harvard.
The COVID-19 Public Projections are intended to function as a resource for very first responders in health care, the public sector, and other afflicted companies preparing for what lies ahead, Google states. They enable for targeted screening and public health interventions on a county-by-county basis, in theory boosting the ability of those who utilize them to react to the rapidly progressing COVID-19 pandemic. Health care suppliers could include the forecasted number of cases as a datapoint in resource preparation for PPE, staffing, and scheduling. Meanwhile, state and county health departments could utilize the projection of infections to assist inform testing strategies and recognize locations at risk of outbreaks.
To produce the COVID-19 Public Projections, Google states its scientists developed a unique time-series maker discovering a technique that combines AI with a creative epidemiological structure. By design, the designs are trained on public information and take advantage of an architecture that allows scientists to dive into relationships the models have identified and interpret why they ensure projections. They’ve likewise been examined to guarantee predictions with respect to people of color —– who have been hardest struck by COVID-19, with disproportionately high rates of cases and deaths —– aren’t hugely manipulated or otherwise deceptive.
“ We observe that our models produce meaningfully lower absolute error and normalized (relative) mistake as compared to the comparison model throughout mainly African American, Hispanic, and white counties,” Google scientists wrote in a fairness analysis of the COVID-19 prediction designs. “ Our models enhance for high precision across all U.S. counties to offer the finest general projection for many communities.”
The COVID-19 Public Forecasts are free to query in BigQuery as part of the service’s 1TB-per-month totally free tier or to download as comma-separated value files (CSVs). Furthermore, they’re readily available through Google’s Data Studio control panel and its National Response Portal.
All bytes processed in inquiries versus the information set will be zeroed out, Google says, but data joined with the information set will be billed at the normal rate to prevent abuse. After September 15, queries over the projection sets will revert to the regular Google Cloud billing rate.
The release of the COVID-19 Public Projections follows the launch of Google’s COVID-19 Public Datasets program, which hosts a repository of public information sets associating with the crisis and makes them much easier to access and examine. Corpora within the COVID-19 Public Datasets program includes the Johns Hopkins Center for Systems Science and Engineering (JHU CSSE) information set, Global Health Data from the World Bank, and OpenStreetMap information, all of which are stored for at no expense on Google Cloud.