Flu season has arrived and all eyes will be turning to Google’s Flu prediction tool, which warns where the flu will hit next by watching people’s internet searches.
Google Flu was hailed as the shining example of how big data could be put to use for the public good when Google first launched it in 2008.
And it worked well back then. Google compared its predictions against the cases that the Centres for Disease Control (CDC) confirmed and its predictions were pretty darn close.
But something went horribly wrong over the past two years, with Google over predicting the number of flu cases by 75% to 95%, the Wall Street Journal reports.
The flu season was bad last year, but nearly not as bad as Google thought it would be.
Essentially what happened was this, Google admitted in a blog post on Friday:
When the flu hit, the media wrote stories about the flu and then people to jumped on Google and start searching about the flu. That caused Google Flu to predict huge outbreaks of the flu which caused the media to write even more stories about the flu, using the Google Flu prediction engine as a source … and so on.
Google also made another mistake with the tool. It only looked at the year-earlier CDC data once. As the CDC started reporting actual cases, Google didn’t keep using the CDC data to check the accuracy of its predictions.
Google says it has updated the tool and will now be using actual CDC data all season long.
The good news is that both Google and the CDC say that flu cases are low so far this season. But it’s early days yet and there are signs that the flu is starting to hit, the CDC says. Last year, the flu hit epidemic levels in some places in January.
The CDC recommends getting a flu vaccine and if you are thinking about getting one, now would be a pretty good time.