model_importance() extracts the variable importance scores from a fitted
workflow for the engines used by the package.
Arguments
- fit
A fitted workflow or model_fit object.
Value
A tibble with a Variable and an Importance column, plus a Sign
column for the (generalized) linear models.
Details
For ranger fits the permutation importance recorded at fit time is returned.
For glm and lm fits the importance is the absolute t- or z-statistic of each
coefficient and the sign is taken from the coefficient estimate, matching the
convention used elsewhere in the package.