This works better if the output of the Ensemble Learning is continuous variable. We take the average of all the weak learners and test on validation data.

• Ensemble model should work better than the individual ones.