Facebook AI predicts SLO County will pass 8,000 COVID cases in 3 weeks. Here’s why
Facebook is predicting that San Luis Obispo County will pass 8,000 coronavirus cases in coming weeks.
Facebook Artificial Intelligence Research, or FAIR, is a part of Facebook’s Data for Good effort.
The Facebook AI research indicates an up to 20-day projection of case increases. Facebook AI updates its forecast every week on the Humanitarian Data Exchange, which allows anyone to access and download the projection data.
Santa Barbara-based nonprofit Direct Relief created a dashboard based on the FAIR project to map out the coronavirus case increases by county.
FAIR’s projection showed San Luis Obispo County would reach 6,198 coronavirus cases by Nov. 30, the Paso Robles Daily News reported. The local case count totaled 6,311 coronavirus cases on Nov. 30.
As of Tuesday, San Luis Obispo County had a total of 6,965 coronavirus cases since mid-March.
The Facebook AI projected that by Dec. 8, the county would be at 6,928 coronavirus cases — 37 coronavirus cases under the actual case count.
Facebook AI projects the county’s case count will increase by nearly 1,000 cases in a 14-day window and reach 7,905 coronavirus cases on Dec. 12.
In just seven days, from Nov. 30 to Dec. 7, San Luis Obispo County added more than 500 coronavirus cases, according to ReadySLO.org. The predicted 14-day increase of nearly 1,000 new coronavirus cases would mean a slight slowdown of transmission.
The Facebook AI forecast reports that San Luis Obispo County will likely reach 8,330 cases on Dec. 28. If the prediction is accurate, that would be a 21% increase in coronavirus cases since Dec. 7.
While Dr. Penny Borenstein, the county’s public health officer said she had not familiar with the Facebook AI model, she said it is likely that the county could reach or pass 8,000 coronavirus cases by Dec. 28 given the “trajectory we have been on over seven to 14 days,” county spokesperson Michelle Shoresman wrote in an email to The Tribune.
According to the Facebook paper detailing the process, the AI model was trained to use time series data about coronavirus spread and uses public data — including the New York Times, Facebook, Data for Good, Google, Carnegie Mellon University’s COVIDcast, The Atlantic’s COVID Tracking Project and the National Oceanic and Atmospheric Association.
“Central to our model is its ability to account for relationships among different counties, so, for example, an uptick in one area can have an impact on predictions for adjacent or similar districts,” a Facebook AI news release read. “This allows us to train models where knowledge about the spread of the disease in one area can improve predictions in a different area.” .
The projections are based on confirmed cases, symptom surveys, community movement, percentage of coronavirus-related doctor visits, testing and weather, according to Facebook.
Facebook AI’s also researched other forecasting models from Columbia University, MIT, John Hopkins and others.
According to Shoresman. San luis Obispo County has been using state models to help make decisions throughout the pandemic.
“Since the beginning of the pandemic, the state (California Department of Public Health) has developed several models by which they are making decisions, including those upon which the recent regional stay-at-home order is based,” Shoresman wrote. “We are generally relying on, and using the state models locally as well, while still looking at our own modeling periodically as needed.”
