by James Harkin

In 2009, engineers at Google claimed they could predict the progress of flu epidemics simply by looking at the words people were typing into their browsers. The idea was both plausible and ingenious; since Google services 3bn search queries every day, the company is quite capable of tracking how often people are asking about flu symptoms or medicine.
Crucially, however, the algorithm for crunching all this information didn’t depend on hunting for obvious flu-related words or phrases. All it needed to do was to work out a correlation – a statistical relationship between previous outbreaks of flu and the search terms being entered at the time. For the engineers at Google Flu Trends these findings, published in the scientific journal Nature, represented a Eureka moment. In a future epidemic, they reckoned, they would be able to chart the spread of the virus as it was happening – and much faster than the medical authorities.

leggi il resto ›