Professor Morteza Karimzadeh said the data helps him and his team measure people's movements and connectivity between locations.
Researchers at the ÀÖ²¥´«Ã½ are using freely availableÌýÌýto forecast COVID cases and hospitalizations for counties and states across the country.ÌýAssistant Professor of GeographyÌýMorteza KarimzadehÌýsaid the data helps him and his teamÌý.Ìý
"These are anonymized aggregate data sets," Karimzadeh said. "Facebook, for instance -- Meta, as it’s called now -- knows how many users it has in each of these counties. It also knows how many of these users are friends with each other on social media. So you can look at this data set and see what proportion of users in these two counties, Denver and Boulder counties, are friends on Facebook. That kind of gives you a measure of how strong the social ties are between these two different counties."Ìý
"We can capture the relationship between counties and states better based on social media friendships. That is just a little mind-blowing to me," Karimzadeh said. "For instance, how connected is Boulder County to Jefferson County and then to Denver County? So if you see a surge in Boulder County or in Denver County, how much of that spills over to Jefferson County?" Ìý
Their predictions, like any other modeling, have a level of uncertainty and probability.Ìý
"We are not fortune tellers," he said with a laugh. "That is why we submit a range for each county and state."
µþ³Ü³ÙÌýtheir estimations can also be really helpful, warning health officials of a potential surge.Ìý
“Initially, I was skeptical that a social media data set would help us improve our results, but to my surprise, it actually helps," Karimzadeh said.
They are one of many teams across the country that submit their forecasts to be collected into a central repository at the University of Massachusetts Amherst. The repository is called theÌý. Agencies like the CDC then use the information from the repository to notify people about virus trends.Ìý
"We can rely on these forecasts and do preventative intervention so the surge doesn't actually happen," Karimzadeh said. "We could essentially put the brakes on the virus before it had a chance to take a toll."
He said the Facebook data sets complement other data they're using to make these predictions, like hospitalizations and deaths.Ìý
While they don't know how long COVID will be with us, they'll continue to harness the power of social media.
"It's one of the biggest data sets that's out there that we can use, and as long as it's out there we will try to tap it as long as it helps with the modeling efforts," Karimzadeh said.Ìý
He said they haven't started using social media data for predicting other diseases yet. But Karimzadeh believes it has potential for forecasting things like the flu.
“If there are warnings of potential surges in the region and health officials advise us to change our behavior for a little while, I think it would be good if we all heeded that advice," he said. “And if it doesn’t happen, then that’s good news. It’s not that the health officials were wrong. It’s just that we prevented that."
See the original original web article on .