Google Flu Trends seeks to measure flu activity in real time by tracking when Google users search for flu-related terms. Developed in 2008, the tool is based on the premise that flu patients will self-diagnose through a Google search before going to see a physician, and would therefore provide Google with a fairly accurate indication of how many people have the flu at any given time, weeks before the Centers for Disease Control and Prevention issues its report on flu activity based on submitted laboratory reports.
However, GFT reported more than twice the number of flu cases last year than were reported to the CDC, according to an article in Science magazine.
"It missed by a huge amount last year and actually, it turns out, it's been missing by a fair amount for several years," David Lazer, PhD, a professor of political science and computer science at Northeastern University in Boston, told NPR. "You could just have used old CDC data — two or three weeks old — and have projected forward, and done a better job than GFT."
So where did Google go wrong?
According to Bloomberg Businessweek, GFT uses 50 million search terms that Google researchers had previously found to be associated with incidences of flu. However, such a large number of terms is bound to find correlation without causation and artificially inflate GFT's numbers.
Google is also constantly adjusting its search algorithms and is more likely to suggest corrected or additional search terms and display advertising more prominently, which may cause users to perform more related searches, according to Bloomberg.
Additionally, a report issued by Google admits an increase in flu-related searches could also have been linked to an increase in media coverage about the flu, prompting healthy users to search for "flu symptoms" and other terms in GFT's algorithm out of curiosity rather than an attempt to self-diagnose.
More Articles on Big Data:
Duke, Thoracic Surgeons Society Partner for Big Data Initiative
HIMSS Announces First Certified Educator for Analytics Maturity Model
Missile Defense Technology Used to Detect Sepsis Earlier