Hey HN,
Every year during the respiratory disease season (winter in the northern hemisphere), the CDC runs challenges where teams -- academics, mostly, but also government and companies -- submit forecasts of the disease burden each week. These are 4-week-ahead forecasts.
We (Emily and I, Joseph) built RespiLens.com as a static website to display these forecasts all in one place with a nicer interface than what is generally provided [1].
Forecasting respiratory disease is challenging, so please interpret forecasts with caution (see how bad it was in the last season).
This website is quite early stage, but we have been using it internally for a year and are starting to build it in public now. We are very much looking for your comments, especially as most folk who have seen it/use it are in the public-health space.
We are not SWEs, so there might be a ton we can improve on the website.
There is also the RespiLens.com/Forecastle, a Wordle-type game to see how good you are at forecasting vs the current state of the art. Interested in what you think of that.
Technically, our GitHub repo is here https://github.com/ACCIDDA/RespiLens, and it's quite straightforward: * we pull forecasts from each hub GitHub; this is facilitated by the HubVerse, an initiative to standardize these challenges (https://hubverse.io) * we serve this with a Mantine Web App. Claude Code and other LLMs were heavily used to create the front-end. It is a good use case because we can do QA as know what it should look like as I have dozens of Python scripts for these plots.
Feel free to ask questions! or suggest features.
[1] hubverse provides an automatic dashboard, and the CDC displays forecasts on their website under a set of conditions (good coverage). See RespiLens.com info hub for link.
Best, Joseph
wosk•1h ago
* Does a comparison with last year help (flu activity is 10% more than last year)? * Should we remove the median from the display? * How hard should it be to check multiple models?
Otherwise, this is a preset link to see the ensemble (the consensus of most models) and our model (influpaint, a diffusion model) and flusion (avery good model by UMass) performance last year:
https://www.respilens.com/?view=flu_projs&flu_dates=2024-11-...