1. A machine learning model is trained on predictions of multiple machine learning models
2. A Logistic regression will definitely work better in the second stage as compared to other classification methods
3. First stage models are trained on full / partial feature space of training data