Communications of the ACM
An Early Flu Season Thwarts Cdc's Forecasting Challenge
This year's influenza season is arriving earlier than usual, with the U.S. Centers for Disease Control and Prevention (CDC) noting the virus is widespread in 12 states, including California and New York. For the last five years, the CDC has hosted a flu forecasting challenge in which academic and industry participants organize predictive models that are assessed at the end of each season. The 35 models for this year are about evenly split, with about half expecting a peak in the last week of the year while the rest are predicting spikes in January and February.
Matt Biggerstaff of the CDC says machine-learning models are mostly limited to near-term projections due to a lack of data with which to train them.
This will be the first year the CDC is supplying state-level data to modelers. Carnegie Mellon University's Roni Rosenfeld says county- or city-level data should be added.
From Wired
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found