Can Twitter and Facebook be used to predict flu (and other) outbreaks?

Researchers in the UK believe they may be able to use Twitter, Facebook, and other real-time social media to predict future events, such as the outbreak of epidemics.

Writing in the journal, ACM Transaction on Intelligent Systems and Technology, scientists from Bristol University said that by collecting Twitter and Facebook updates over a number of months they could monitor outbreaks as they happened.

Professor Nello Cristianini  told the Press Association that the research group was then able to use this data to estimate the scale and growth of flu outbreaks in specific areas into the future.

The scientists gathered over 50 million geo-tagged tweets and analysed the text for any mentions of terms related to flu, they then compared this data with information from the UK’s National Health Service to measure the accuracy of their methodology.

Professor Cristianini said “Our research has demonstrated a method, by using the content of Twitter, to track an event, when it occurs and the scale of it.

“We were able to turn geo-tagged user posts on the microblogging service of Twitter to topic-specific geo-located signals by selecting textual features that showed the content and understanding of the text.”

This isn’t the only attempt to use the web to predict flu outbreaks, since 2007 Google’s Flu Trends website has been tracking the number of searches with terms related to flu in order to predict and track the outbreak of flu on a global scale in near real-time.

Google says that by tracking known key terms (and accounting for other ‘noisy’ searches, such as for drug recalls) its Flu Trends website can see “strong historical correlations” with known flu outbreaks

View Larger Map

Of course this methodology doesn’t just have to be used for predicting outbreaks of disease. It could be used to predict where protests are likely to occur; areas suffering from flooding after heavy rain; or, as some suggest, even predict the stock market and election results.