Adds a new utility function for filtering a table based on importance scores, quartiles, or feature counts. This function is used in plot_feature_volatility and exposed in that action as well as its parent pipeline, feature_volatility. Default is to filter 100 most important features, so this should fix the problems described in #144.
Adds new unit tests for utility described in item 1.
Adds param missing_samples to error or ignore missing samples in plot_feature_volatility. (this is not in response to an issue nor is it necessarily a bug — this action is usually run as part of a pipeline where this will not be an issue — just being proactive).
fixes #144
This PR:
plot_feature_volatility
and exposed in that action as well as its parent pipeline,feature_volatility
. Default is to filter 100 most important features, so this should fix the problems described in #144.missing_samples
to error or ignore missing samples inplot_feature_volatility
. (this is not in response to an issue nor is it necessarily a bug — this action is usually run as part of a pipeline where this will not be an issue — just being proactive).