Once there is a basic model ranking system in place ( https://github.com/dssg/tyra/issues/37 ) we would like to implement some more rankings based on the selected metrics and allow them to be switched between. Examples:
a slope close to 0
variance closest to 0
the highest minimum value
the highest maximum value
We will likely want to include some kind of training explanations for what these mean, once we are referencing them in the UI. This may even be encoded in the name; for instance, 'slope closest to 0' could mean 'most stable over time', with a more detailed explanation available as an info box.
Once there is a basic model ranking system in place ( https://github.com/dssg/tyra/issues/37 ) we would like to implement some more rankings based on the selected metrics and allow them to be switched between. Examples:
We will likely want to include some kind of training explanations for what these mean, once we are referencing them in the UI. This may even be encoded in the name; for instance, 'slope closest to 0' could mean 'most stable over time', with a more detailed explanation available as an info box.