Sickweather AI accurately predicted late flu outbreak

Tools

Source: Sickweather

According to the CDC, the annual flu season is between October and May and it commonly peaks between December and February.  This year, there's a late flu outbreak, and in January, Sickweather's Nightingale, an emerging AI forecasting engine, accurately forecast its appearance in March. And here it is, right on cue.

Not that anyone wanted that prediction to be correct, because who wants to worry about catching the flu now? Nonetheless, forewarned is forearmed especially if you take illness predictions seriously and bare your arm for a vaccine.

"We're very encouraged by these results," said Graham Dodge, CEO of Sickweather, in a statement to the press.  

"Our data science team has been testing this forecast model since August, and while results are still coming in, I think we nailed it."

Predictive analytics have many uses, but illness predictions land somewhere near the top of the most wanted list. That's why Google Flu Trends and the CDC have been working on using predictive analytics to nail down flu danger periods in advance for quite some time now.

The CDC bases its predictions on healthcare professionals reporting and good-old fashioned surveys. Google Flu Trends initially relied on big data analysis of search and social media data to make its predictions. But after overshooting the mark, Google Flu Trends announced it would also use CDC data in the analysis.

By comparison, Sickweather provides a real-time map of illness by analyzing data from public reports of illnesses posted on social media and by direct reports made through their free apps.

It will take time to figure out which data is most essential to accuracy in illness tracking and predictions. Meanwhile, it's great to see so many approaches tried and the short-term results from those efforts. Each adds to our knowledge base and that's a very good thing.

Related Articles:
Google Flu Trends gets a data makeover
Google vs CDC: When traditional methods beat big data