Meet Stacker — automatic python ensembling factory.

This is Part II of the series where we will explore a fully automated highly configurable models stacking framework

If you need to refresh the concept of stacking, refer to Part I which goes over the theory, explains all the tips&tricks, provides benchmarks on the selected Kaggle competition and features code examples of how to create an ensemble manually

Update 12/2022: as a feature request an additional set of functions to save fitted stacker instance and load it from disc has been introduced. Refer to the closing remarks of this article and the documentation.

Jumping ahead:

The GIF above just showcased verstack.Stacker in action. A user has initiated the Stacker class, defined which models they want to stack in each layer and executed stacking of those layers by creating additional features in their train/test sets