nssac / cdc-flusight-ensemble

Guidelines and forecasts for a collaborative U.S. influenza forecasting project.

This repository is the central nervous system for the FluSight Network, a consortium of researchers and public health officials working together to create real-time forecasts for influenza in the US.

There is a lot of data and code in this repository. We have over 20 models' forecasts for the past nine influenza seasons in the US. To help you navigate it, here are some quick links to relevant information and files.

Useful Links

• This collaboration is currently centered around participation in the CDC FluSight challenge which has run since the 2013/2014 season.
• The forecast data in the repository is made available under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0). Note that the manuscript data in the repository may have different licensing, due to publisher's licensing.
• In the fall of 2017, Nick wrote a blog post describing the new FluSight Network effort effort. This is a good place to start to get an overview of this collaborative effort.
• The first academic paper resulting from this collaboration was published in January 2019 in PNAS. This paper compares performance of the models over the 2010/2011 through 2016/2017 flu seasons. The source code for manuscript itself also lives in this repository.
• The second academic paper resulting from this collaboration is available as a pre-print. This paper describes our work to build a collaborative multi-model ensemble in the 2017/2018 influenza season.
• We have an interactive visualization of the forecasts available at flusightnetwork.io.
• The technical details and structure of this repository is described on the wiki for this repository.
• We have specific guidelines for how to join the FluSight Network and participate in the current season.
• We are working on a second manuscript with details on our building a collaborative ensemble.
• The Reich Lab is building a forecast repository, Zoltar, which boasts a standardized API access and visualization framework. The forecasts from this FluSight Network project, in addition to living here on github, also have been stored in a Zoltar project.

cdc-flusight-ensemble's Issues

Debugging onset prediction Region 1, 2013-14

Onset predictions for Region 1 in 2013-14 seem to swing wildly between weeks 3-6 and 'no onset'. Not sure why, and it might be related to the similar wild swings in onset scores overall. Would be good to fix this at least in this particular case. @christa60

Adding 2014EW10 runs for Region 9

Seems EW10-2014-FluX_LSTM.csv is missing Region 9's forecasts (most likely due to a failed rivanna job). @christa60 could you check, rerun and update?

Thanks,
Srini