Building a reduced model
Variable importance analysis helped you identify the most predictive features from the attrition
dataset. Based on it, you will build a drastically reduced model with only three variables: OverTime
, DistanceFromHome
, and NumCompaniesWorked
and compare its performance to the full model baseline. The metrics you estimated for the full model are stored in aug_full
.
All data, along with the train
and test splits, is available in your environment.
This exercise is part of the course
Feature Engineering in R
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Create a recipe using the formula syntax that includes only OverTime, DistanceFromHome and NumCompaniesWorked as predictors
recipe_reduced <-
___(Attrition ~ ___ + ___ + ___, data = train)
# Bundle the recipe with your model
workflow_reduced <-
workflow() %>%
add_model(model) %>%
___