Video by freeCodeCamp.org · taught by Vincent D. Warmerdam · streamed from YouTube. All rights remain with the creator.
Lesson 5: Scikit-learn Pipelines, Metrics and Meta Estimators
About this lesson
This crash course by Vincent D. Warmerdam is aimed at people who already know the basics of machine learning and want to use scikit-learn the way experienced practitioners do. Rather than surveying algorithms one by one, it concentrates on the design ideas that make the library powerful and the habits that lead to trustworthy models. The overview section introduces scikit-learn and its high-level building blocks, showing how models, transformers and pipelines fit together and how settings are searched systematically to find good ones. The section on preprocessing tools then demonstrates how transforming the data before it reaches the model can affect performance as much as the choice of algorithm itself, with examples that compare the results of different transformations. Metrics come next. You learn how to judge models with the scores the library provides, why the right metric depends on what a mistake costs, and how to create custom metrics of your own when the built-in ones do not capture what matters for a problem. The section on meta estimators covers tools that wrap ordinary models to change or extend their behaviour, including ways to post-process predictions. The course ends by introducing a separate library that builds on scikit-learn and aims to make machine learning more approachable by adding extra components that plug straight into pipelines. Afterwards, read Chapter 5 of the course book, which works precision and recall across thresholds by hand.