The FluSight Network

Collaborative, multi-team, ensemble influenza forecasting

Overview

Since the 2013/2014 influenza season in the US, the CDC has organized the FluSight challenge. Teams from around the globe submit weekly forecasts about the current and projected trajectory of the influenza season. The Reich Lab began participating in the FluSight challenge in the 2015/2016 season. In March of 2017, we led the effort to establish the FluSight Network, a consortium of teams participating in the FluSight challenge, with the goal of creating an ensemble that weighted models differently depending on their past performance. Working with teams from Carnegie Mellon, Columbia University, Los Alamos National Laboratory, and, in 2018/2019, Protea Analytics, we developed and implemented the FluSight Network Ensemble Model. This model was the second best overall performing model in the FluSight challenge during the 2017/2018 season. Based on its strong performance in this pilot season, the CDC chose to use the forecasts from this model on its official FluSight webpage during the 2018/2019 influenza season.

Links