Recruiting the base learners

Ensemble methods use multiple learning algorithms to obtain better predictive performance than what any of them alone could have achieved. That being said, it is not true that combining multiple learning algorithms always leads to a superior result.

In this series of exercises, you will predict whether a star belongs to the pulsar class given eight numerical attributes. The decision attribute is named Class and takes yes/no values. You will train four learning algorithms on identical resamples of the training data and check their individual performance. Then you will check if they are suitable for building an ensemble and apply the stacking strategy.

The caret and caretEnsemble packages have been preloaded and the training dataset is available in your workspace.

Question

What is an important aspect to consider when choosing the set of base learners for an ensemble model?

Possible Answers

That they are weakly correlated with each other in their predictions.
That they all belong to the same family of algorithms (e.g., decision trees).
That they have low computational complexity.
That they are robust to outliers and missing data.

Exercise

Recruiting the base learners

Instructions 1/4

Question

Possible Answers