Started on a SHAP plot. This is a popular method for feature importance that's more robust than the ones built in to sklearn. It provides an example of a plot that needs to be saved as an image since it would be hard to support as a vega plot.
Example output:
One downside is that it's expensive, so it might make sense to sample the data here, but I didn't bother for now.
dberenbaum
commented
1 year ago
Started on a SHAP plot. This is a popular method for feature importance that's more robust than the ones built in to sklearn. It provides an example of a plot that needs to be saved as an image since it would be hard to support as a vega plot.
Example output:
One downside is that it's expensive, so it might make sense to sample the data here, but I didn't bother for now.