It turns out that complaining on social media about how sick and miserable you feel might be useful after all!
A new social media research project claims that it's possible to track some flu epidemics just by watching trends on Twitter. As reported by the MIT Technology Review, the work of Jiwei Li at Carnegie Mellon and Claire Cardie at Cornell University aims to detect early outbreaks of influenza by monitoring the location and proliferation of flu-related Tweets on the social network.
Waaaa I have flu. Watching friends on E4 at 11am and crying into berrocca 😨 🙈— Foxes (@iamfoxes) October 11, 2013
“Based on the Twitter API, we developed a crawler to fetch data at regular time intervals,” the team wrote in the published research. “We fetched tweets containing indicator words [flu, influenza, H5N1, H5N9, swine flu, bird flu] and collected 3.6 million flu-related tweets from 0.9 million twitter users starting from Jun. 2008 to Jun. 2010. Location details can be obtained from the profile page or mobile client.”
A similar project, Google Flu Trends, was launched by Google's philanthropic and charitable arm, Google.org, in 2008, with its predictions coming from analyzing worldwide flu-related searches and their locations.
Both systems by Google and by Li and Cardie mapped out data to check against historical CDC statistics, with the Twitter model showing a bit more correlation to recorded cases of the flu. The next step would be to use the model as a projection to help produce an even earlier system of outbreak prediction than what already exists.
“We test our model on real time datasets for multiple applications and experiments results demonstrate the effectiveness of our model. We are hopeful that our approach would help to facilitate timely action by government and health organizations who want to decrease the number and cost of unnecessary illnesses and deaths,” the team concludes.
Since its popularity explosion several years ago, Twitter has become a fairly popular research tool for social scientists and research groups.
For example, one project by the University of Vermont, reported by ABC News in May, tracked the “happiness” of tweets, with the death of Michael Jackson, the Japanese tsunami in 2011 and the 2012 school shootings in Newtown, Conn., as being some of the saddest days on Twitter since 2008. The ongoing study measures the use of positively valued keywords such as “happy,” “merry” and “family,” and negatively valued keywords such as “bomb,” “sad” and “prayer.” Twitter itself is also experimenting with data analysis to inform users of Twitter accounts that friends have followed, and of breaking news events of interest.
So, remember, Twitter isn't just for witty one-liners and linking to blogs anymore. It's for science.