Researchers from Johns Hopkins University in Baltimore and George Washington University in Washington, D.C., have developed an algorithm using data from Twitter that is able to predict flu outbreaks with 85 percent accuracy.
To develop the algorithm, researchers used a sample of 10,000 tweets, identifying those that referenced a compliant about flu symptoms. Researchers then used the characteristics and phrases from these tweets to develop an algorithm able to identify relevant tweets and point to areas of flu outbreaks based on the users' locations.
When compared with data from the Centers for Disease Control and Prevention from the 2012-2013 flu season, the algorithm's results were 93 percent accurate. However, the main benefit of the algorithm is the ability to identify potential outbreaks in real time — when analyzed retrospectively, the algorithm was able to predict weekly changes in flu levels with 85 percent accuracy.
Having real-time information about flu trends locally can help hospitals prepare for looming outbreaks. "We’re actually able to track at the municipal level," David Broniatowski, PhD, an assistant professor in the George Washington University's Department of Engineering Management and Systems Engineering, told the Journal. "We can provide them with the data about what the flu is like in their city. That allows them to do surge planning."
A similar project, Google's Flu Tracker, was recently shown to be inaccurate, as it predicted more than twice the number of flu cases last year than were reported to the CDC for the 2012-2013 flu season. The researchers working on the Twitter algorithm sought to avoid a shortcoming of Google's method — counting users who were searching for information about the flu but were unlikely to have symptoms — by using Amazon's Mechanical Turk service to hire freelancers to identify which tweets of the original sample of 10,000 were likely to be posted by someone with the flu rather than someone simply discussing the flu, according to the Journal.
More Articles on Big Data:
UC Davis Researchers Develop Sepsis Detection Algorithm
Big Data Lessons From Google's Inaccurate Flu Tracker
Duke, Thoracic Surgeons Society Partner for Big Data Initiative