Predicting Flu from the Content of Twitter

Social networking is used by hundreds of millions of people to exchange personal information. Online friendships, virtual groups, business oriented communities, entertainment, almost everything is available on this part of the web. In social networks such as Twitter, user posts are disclosed publicly by default providing a rich set of domains to infer interesting aspects of our everyday life.

Researchers from the recently formed Intelligent Systems Laboratory are exploiting this fact. Vasileios Lampos, a PhD student, his advisor Prof. Nello Cristianini and Dr Tijl De Bie in their latest work use tweets geolocated in the UK to infer the level of flu-like illness in the population.

Their methodology is drawn from the broad scientific field of Artificial Intelligence. Statistical Machine Learning techniques are applied to form predictions for an online tool: the Flu detector. The usefulness of the work lies on the fact that these predictions are made available daily providing inexpensive and timely information about the state of an epidemic. Technical details can be found in their recent publications.