Variable importance
You already know that bagged trees are an ensemble model that overcomes the variance problem of decision trees. Now you learned that the random forest algorithm further improves this by using only a random subset of the features in each tree. This further decorrelates the ensemble, improving its predictive performance.
In this exercise, you will build a random forest yourself and plot the importance of the predictors using the vip
package. The training data, customers_train
, is pre-loaded in your workspace.
This exercise is part of the course
Machine Learning with Tree-Based Models in R
Exercise instructions
- Create
spec
, the specification of a random forest classification model using the"ranger"
engine and"impurity"
variable importance. - Create
model
by fitting the tibblecustomers_train
tospec
usingstill_customer
as the outcome and all other columns as the predictor variables. - Plot the variable importance using the
vip()
function from thevip
package (which is not pre-loaded).
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Specify a random forest
spec <- ___ %>%
set_mode("classification") %>%
set_engine(___, importance = ___)
# Train the forest
model <- spec %>%
fit(___,
___)
# Plot the variable importance
vip::___(model)