In collaboration with the Harvard Health Institute, the Google today released its Public Forecasts for COVID-19, a set of models that provide projections of COVID-19 cases, deaths, ICU utilization, ventilator availability and other metrics for the next 14 days for US 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 will continue to be updated with guidance from its Harvard partners.
Public forecasts for COVID-19 (COVID-19 Public Forecasts) will serve as resources for first-door scientists in healthcare, the public sector and other affected organizations preparing for the future.
They enable targeted trials and interventions in public health from county to county, theoretically enhancing the ability of data users to respond to the rapidly evolving COVID-19 pandemic.
For example, healthcare providers could integrate the envisaged number of cases as a reference point in their resource planning, staffing and service scheduling.
Meanwhile, state and county agencies could use infection forecasting to conduct strategic tests to identify areas at risk.
To create public forecasts for him COVID-19, Google reports that its researchers have developed a new time-series machine learning approach that combines AI with an intelligent epidemiological base.
From their design, the models are trained in public data and utilize an architecture that allows researchers to find relationships they have identified (models) to interpret them because they make specific predictions.
They are also programmed to ensure the predictions in relation to countries who have been most affected by COVID-19, for there is a higher success rate, due to the larger number of data.
