In-sample and out-of-sample performance
Does a more sophisticated model always perform better? As we discussed in the video, that's only half the truth.
Overfitted models understand the structure of their training set perfectly but cannot generalize to new data. That's a bummer! At the end of the day, the main purpose of a predictive model is to perform well on new data, right? Go investigate!
Pre-loaded is the last model of the previous exercise, complex_model
, and your training and test data (chocolate_train
and chocolate_test
).
This exercise is part of the course
Machine Learning with Tree-Based Models in R
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Predict on and combine with training data and calculate the error
predict(___, new_data = ___) %>%
___ %>%
mae(___,
___)