COVID-19 research: Geographer tracks movement with Twitter data
Study holds promise for disease tracking in future outbreaks
Posted on: April 7, 2020; Updated on: April 7, 2020
By Chris Horn, chorn@sc.edu, 803-777-3687
During historic flooding five years ago across the Palmetto State, faculty members at the University of South Carolina demonstrated how real-time social media data could aid in tracking the extent of a natural disaster.
Geography assistant professor Zhenlong Li led that research and has deployed similar methods during the current COVID-19 pandemic, gathering Twitter data to visually map human travel across the country and around the world. When it is better refined, Li says, the technique could aid public health officials in predicting the spread of future disease outbreaks or in other crises where human movement is a critical factor.
“Monitoring human movement at different geographic scales (from global to local) is critical for us to gain better understanding on the spatial spread of the virus,” Li says. “These initial maps were just to gauge interest and demonstrate what we can do. The most important question is how to build predictive models incorporating such near real-time population flows for better infectious risk prediction.”
That sort of fine-grained analysis could help public health officials to estimate the risk of disease outbreak in a specific location.
“Integrating the movement information we’ve captured with Twitter, trends in daily infection cases and socioeconomic data, our next step is to use artificial intelligence to predict current and future infectious risk at the county and state levels,” Li says.
Analysis of human movement patterns over space and time could also help public health officials to assess compliance and effectiveness of shelter-in-place and other public safety orders, Li says.
Li acknowledges there is bias built into the Twitter data because Twitter users tend to skew toward a younger demographic. Manually checking Twitter users’ profiles can provide a more accurate estimate of age but would be prohibitively time consuming with large datasets. Using AI to mine such information would be more efficient but that may trigger privacy concerns, Li says.
Still, Li says this type of geospatial research could prove useful during future influenza outbreaks and other health crises.
Share this Story! Let friends in your social network know what you are reading about