This presentation introduces the scikit-learn library. It describes the three main types of objects formally and operationally: transformer, estimators, and predictors. Particular attention is paid to the GridSearchCv and Pipeline objects. They are widely used in data preparation and model training.
The attached Jupiter Notebook shows a standard application of all objects presented in a regression problem. By taking advantage of the formal definitions provided, I show how to build custom objects. They can be integrated into scikit-learn, in case of more specific needs.