Twitter data stream, effective in preventing flu

Google launched the Flu Trends website back in 2008. He said that people who suffer from illness can spend more time searching for disease-related content than when they are healthy. Therefore, by analyzing the number of flu-related searches in a given country over a given period, the spread of the virus can be extrapolated.

Of course, this assumption is very valid and the projections are very reliable, often with little difference from the data from the Centers for Disease Control and Prevention (CDC). It is well-known that the initial prediction of the disease will provide a favorable opportunity for the government to make timely deployment. The timing of online predictions is much earlier than the forecast for the early stages of an infectious disease.

So can other online services, especially social networks, achieve the same or better predictions? Today's questions are answered. Jiwei Li of Carnegie Mellon University and Claire Cardie of Cornell University used Twitter successfully to predict early flu outbreaks.

Their way is very similar to that of Google. First, they filter tweets that contain "flu," with positional labels, from Twitter data streams; then, map the distribution of those tweets on the map, and the changes over time. At the same time, they also produced a dynamic model of the flu. In the new model, the flu consists of four stages: the non-contagious phase, the outbreak phase, the stabilization phase and the declining phase.

In addition, they adopted a completely new algorithm to try to find switching nodes of different periods as fast as possible. In fact, Li and Cardie validated the effectiveness of the method with 3.6 million tweets by 1 million Americans between June 2008 and June 2010. In order to test whether their prediction is true, Li and Cardie compare their analysis with CDC. They said, "We are convinced that there is a significant correlation between flu-related tweets and the number of flu cases provided by the CDC."

This appears to be a powerful weapon for declaring war on the flu. Not only does it provide a new way for early disease prediction, it also challenges the CDC and Google's flu site. Every year, about 10-15% of the world's people are infected with flu, causing 50 million flu cases, killing nearly 500,000 people. This is a huge number that can not be ignored. Nowadays, governments and medical institutions can use Twitter to predict the outbreak earlier and more affordable and more conveniently to save more lives.

Original link: http://www.36kr.com/p/206868.html

【Editor's Choice】

Twitter stocks death squads: data woven into the money and look calmly deal with 143,000 TPS Twitter is how to make it! Twitter launches Storm-Hadoop hybrid Twitter with its profitable data business 【Editor: Xiao Yun TEL: (010) 68476606】