With flu season just around the corner, Google has updated the Google Flu Trends service to include digital data from its search engines and traditional data to provide a more accurate prediction of flu outbreaks around the world.
Google isn't turning a deaf ear to criticisms of its Google Flu Trends prediction tool. The service was launched in 2008 aiming to take advantage of big data and inform healthcare workers and the general public about the intensity of the flu season in 29 countries ahead of official announcements.
The predictions, which were based entirely on the number of flu-related searches on Google's search engine, were mostly accurate during the tool's early days, even correctly predicting the severity of H1N1 or swine flu two weeks before official statistics came out in 2009.
However, by 2012, Google Flu Trends' predictions became increasingly off the mark. Computer science professor David Lazer of Northeastern University tells the Wall Street Journal that Google overpredicted flu cases 95 percent of the time that year.
Although Google tweaked its algorithm in 2013, the prediction tool still overshot its calculations 75 percent of the time.
Lazer, who published an article on Google Flu Trends with a team of three quantitative social scientists, says analysis of the prediction tool shows a combination of digital data gleaned from Google search and traditional data from the Centers for Disease Control and Prevention (CDC), which gathers data from around 3,000 healthcare providers, served the most accurate prediction of flu trends.
The updated Google Flu Trends will now be offering data from this combination of sources, although Google has declined to provide details into what exactly has changed.
In a blog post, Google senior software engineer Christian Stefansen says Google.org, which handles Google Flu Trends, will be publishing a technical paper on the matter, but it will not include the search terms that researchers like Lazer and his team would like to see.
Stefansen tells the Journal that they "would love to" provide the search terms to independent researchers, "but if we were to do that, it would be easy for someone to game the system."
The specifics of how Google's search engine works have always been a mystery to competitors, marketers and advertisers who want to know how Google defines what makes it to the top of its search results. Providing the search terms that define its flu predictions would give the public a clue into how Google's search technology works.
"We're at this intersection between providing a service for free and making it researchable, so we're trying to strike the best of both worlds," Stefansen says.
