Algorithm

• used to create an ensemble from a diverse group of learners
• Rather then simply using a majority vote or median output of the group of learners,
• a meta-learner is trained to learn how to combine the output of the individual learners
  • trained using the prediction made by the base models, and not the raw-data
  • The meta-learner is trained on a data set that was not used to train the diverse group of learners.

Algorithm

(1) Train a user-specified number of base learners (2) Generate an out-of-sample prediction for each data instance using each of the base learners (3) Train the meta-learner on the predictions generated in step 2 and their associated target values