Tracking the Flu Through Twitter
Johns Hopkins researchers go local with their Twitter flu tracking efforts
Researchers at Johns Hopkins University have begun using Twitter data to track the spread of the flu virus in both national and local settings. They’ve determined that Twitter can be an incredibly useful analytical tool in gauging the spread of the flu and predicting trends. This information can be used to prepare adequate medical care ahead of time and prevent the spread of the flu before it advances too much. By preparing hospital beds and vaccinations in at-risk areas, the spread and damage of the flue can be reduced significantly.
The spread of the flu can be represented as a graph where edges represent an infection from one person to another. The theory of networks can be used in various ways to help predict the form that this graph will take. It can likely be inferred that a person with many followers on Twitter, or a person that tweets or is retweeted a lot, is more sociable in reality. This person will be likely to have more strong and weak ties, and thus be more likely to spread the disease.
Furthermore, if a person in a currently infected person’s twitter followers becomes sick, it may be that the disease was transferred from the first person. By constructing a model based on the degrees of separation from one person to another with a direct line establishing infections between them, one could accurately predict how far the disease has spread and how far it is likely to spread in the future. Six degrees of separation can seem very small when one considers that it takes just six steps from patient zero to infect the entire world, but of course in reality not everyone that a person contacts will become infected or spread the infection to others. By using this model, one can find dense sections of the Twitter follower graph where people are separated by very few degrees, and use this knowledge to predict how quickly the flu will spread through any given population. These dense sections would generally represent friend groups, but could also represent coworkers and other people that would be likely to interact with each other frequently.
Research like this has much potential for expansion; if given enough data from other social media such as Facebook, researchers could build an accurate and expansive connected graph to accurately predict the spread of the flu and of much more deadly diseases within any population. Such a project has the potential to save lives with quick intervention and prevention, and is definitely worth further consideration.