Twitter Used to Predict Health Patterns
The University of Rochester revealed on Friday that Twitter can be used to predict health patterns and provide a better understanding of the influence of lifestyles. The research relies on a combination of several factors including GPS tags in statuses to find patterns on Twitter. This information is being used to understand overall health in different regions of the world.
Researchers are gathering information from Twitter and mapping it to show how lifestyles impact health. Factors such as pollution and activity levels are being used to see the influences on users’ health. The study was presented at the International Conference on Web Searching and Data Mining in Rome on Feb. 8.
The information presented by the University of Rochester has been used to create a GermTracker that shows the frequency of certain illnesses on a map, and data from users’ Twitter accounts is essential for building the tracker. Some interesting patterns have emerged from the data including how daily social interactions influence health. For example, users who spend more time going out to their local gyms are more likely to get a cold or flu than those who stay home.
Combining tweets to search for patterns requires an algorithm that is capable of differentiating between passive posts about health and real instances of people being sick. The danger of getting diluted results that influence data is real. In addition, the number of celebrities and athletes who post on Twitter can affect results because of the large number of retweets. For example, posts such as Shaquille O’Neal’s promotion of Icy Hot and Jason Kidd mentioning his use of Heel That Pain have to be considered. It is not an easy task to tell the difference between sympathy tweets for Justin Bieber after he announces having a cold and real posts about being sick. Researchers admit that finding the right algorithm is an ongoing process.