Preprocess
Feature engineering time! You need to build a recipe to take care of non-informative but possibly valuable variables such as observation ID or deal with missing values. This is also an opportunity to transform some predictors. Say, normalize numerical features and create dummy variables for categorical ones.
The attrition
dataset and the train
and test
splits you created in the previous exercise are available in your environment.
This exercise is part of the course
Feature Engineering in R
Exercise instructions
- Normalize all numeric features.
- Impute missing values using the
knn
imputation algorithm. - Create dummy variables for all nominal predictors.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
recipe <- recipe(Attrition ~ ., data = train) %>%
update_role(...1, new_role = "ID") %>%
# Normalize all numeric features
___(all_numeric_predictors()) %>%
# Impute missing values using the knn imputation algorithm
___(all_predictors()) %>%
# Create dummy variables for all nominal predictors
___(all_nominal_predictors())
recipe