Summary

In this chapter, we looked at the very important machine learning methods of creating an ensemble model by stacking and then multiclass classification. In stacking, we used base models (learners) to create predicted probabilities that were used on input features to another model (super learner) to make our final predictions. Indeed, the stacked method showed slight improvement over the individual base models. As for multiclass methods, we worked on using a multiclass classifier as well as taking a binary classification method and applying it to a multiclass problem using the one-versus-all technique. As a side task, we also incorporated two sampling techniques (upsampling and Synthetic Minority Oversampling Technique) to balance the classes. Also significant was the utilization of two very powerful R packages, caretEnsemble and mlr. These methods and packages are powerful additions to an R machine learning practitioner.

Up next, we are going to delve into the world of time series...