We could use scikit-learn or other machine learning libraries with accepted and robust implementations to compare, how our predictions perform.