Facebook AI predicts surge of COVID-19 cases in SLO County
Map of predicted COVID-19 case spread by county, based on Facebook AI Research. Click here for map with predictions.
–With COVID-19 case numbers spiking in California and across the country, new restrictions are being put in place at the city, county, and state levels to help address the pandemic. Last month, Facebook announced that it is now using a computer system with artificial intelligence (AI) to forecast the spread of coronavirus across the entire United States. It is part of Facebook’s Data for Good initiative.
Unlike existing insights which focus on state-level predictions or death counts, this forecast uses anonymized Facebook data to deliver granular county-level results.
The AI projects that San Luis Obispo County will see 6,198 cases of COVID-19 by Nov. 30, which represents an increase of almost 900 cases in two weeks, a 17-percent increase since the pandemic began in March. The AI predicts more than 2.5 million new cases in the US by the end of the month. See map of predictions by county here.
The AI predicted there would be 5,420 cases of COVID-19 in the county on Thursday. San Luis Obispo County Health Agency officials reported an actual total of 5,486.
“Here in California we are seeing increased rates and this data is vital to understanding how the outbreak is spreading,” says Janet Fernandez with Elevate Public Affairs. “This can help decision makers across public health systems make sure they’re sending resources to the right areas, but getting good data at the county-level can be challenging. Facebook hopes these new forecasts help public health experts and local authorities understand where prevention efforts are working and which areas are expected to be hotspots.”
Facebook said in an announcement, “AI is an important tool to support public health experts around the world in their efforts to keep people safe and informed amid the coronavirus pandemic. Facebook AI is partnering with academic researchers and other experts on a range of initiatives related to COVID-19. We are sharing overviews of several of these now, and we will add updates and more information in the days and weeks to come.”
COVID-19 AI prediction data sets include:
- Confirmed cases based on reports from state & local health agencies – The New York Times
- Symptom survey, the prevalence of COVID-like symptoms from self-reported surveys – Facebook Data for Good
- Movement range maps – Facebook Data for Good
- Community mobility, movement trends across different categories of places, such as grocery stores, recreation, pharmacies – Facebook Data for Good
- Doctor visits, percentage of COVID-related doctor’s visits in a given location – CMU COVIDcast
- Testing, the total number of COVID PCR tests per state – The COVID Tracking Project
- Weather, average, minimum, maximum temperature & rainfall per county – NOAA
“To generate these nationwide county-level forecasts, Facebook trained our AI models using time data about the spread of the disease consisting of county names, dates, and number of confirmed cases. We also taught it about possible signals that can influence the spread of the disease, such as mobility, social distancing, and the prevalence of COVID-like symptoms,” a Facebook spokesperson says.
“To improve the accuracy of our forecasts, we also developed a new AI model that disentangles regional from disease-inherent aspects within these data sets. Central to this model is its ability to account for relationships between different counties, so for example, an uptick in one area can impact predictions for adjacent or similar districts.”
Facebook published a report on the model’s methodology here.
