Google announces update to Flu Trends model
Google first built the model to estimate flu trends in 2008 by using aggregated Google search data from regions around the world.
The Flu Trends team evaluates the performance of the model at the end of every flu season to see if adjustments need to be made. At the end of the H1N1 flu season in 2009 updates were made to ensure more accurate estimates.
Flue Trends helps estimate the start, peak and duration of each flu season. During the 2012-2013 season, the model correctly predicted the start and end of the season, but overestimated the number of cases. After investigation, the team found that because more people were searching for information on the Internet than in previous years, their model predicted a higher number of cases.
The team looked at many options for improving the model, and decided to use the 2012-2013 season peak as a close approximation of activity. The team said the update would be applied to flu level estimates for the 2013-2014 flu season starting from Aug. 1.
Flu Trends said that with the update, the model would show a lower forecast than last year's model, but they believe the lower number more closely approximates CDC data.