Toy-project using artificial unbalanced data.
Supervised learning pipeline for multiple classifiers and parameters tuning is performed via RandomizedSearchCV.
In order to ensure that in case of large datasets memory exhaustion is avoided subprocessing is implemented for various combinations of parameters.