An interactive dashboard made with Jupyter and Voila. Users can play around with parameter like class imbalance, model strength or cutoff value and observe the effects on metris like ROC/AUC or accuracy/precision/recall.
Now available live here and in the official Voila Gallery.
Some code copied from this voila example.
Deployment: https://voila.readthedocs.io/en/latest/deploy.html#deployment-on-heroku
- More comprehensive Readme on machine learning model evaluation, describing metrics in detail
- Create GIFs and add to Readme
- Smoother refresh and faster plot update. Could be possible via using matplotlib's notebook or widget backend, but currently not working, see this GitHub issue.