datatable.models.Ftrl.feature_importances¶
feature_importances
Feature importances as calculated during the model training and
normalized to [0; 1]. The normalization is done by dividing
the accumulated feature importances over the maximum value.
Parameters¶
return
FrameA frame with two columns: feature_name that has stype str32,
and feature_importance that has stype float32 or float64
depending on whether the double_precision
option is False or True.