Google Flu Trends was wildly inaccurate last year. Here's how researchers improved it

Last winter, Google Flu Trends was shown to be a work in progress when it wildly overestimated incidence of influenza, but its algorithm-based model has considerable potential. Now, researchers have combined Flu Trends with CDC data and machine learning to improve forecasts. The outcome is a tool that perfectly predicted flu season peaks in several cities up to 9 weeks before they happened. Item