Scikit-learn tutorials that outlived Python 2

What it does

This repository hosts the Jupyter notebooks for a 10-part, 4.5-hour video introduction to scikit-learn. The materials move through the standard supervised-learning workflow—loading the iris dataset, training K-nearest neighbors models, comparing estimators, and finally building preprocessing pipelines with pandas and seaborn. Each notebook is framed as a sequence of concrete questions, so it reads more like a guided conversation than a reference manual.

The interesting bit

The noteworthy detail here is maintenance. The original recordings ran on Python 2.7 and scikit-learn 0.16, but the notebooks have since been rewritten for Python 3.9.1 and scikit-learn 0.23.2. That makes this one of the few video-era tutorials whose companion code has not been left to quietly fossilize. The author even stashed the original versions in an archive branch for historical comparison.

Key highlights

• 10 topic-driven notebooks covering model training, evaluation, cross-validation, grid search, and Pipelines
• Tied to a free YouTube playlist and a separate online course that adds quizzes and a completion certificate
• Bonus PyCon 2016 tutorial on text-based data (vectorization, model comparison) housed in a linked repository
• Explicit question-and-answer structure: each section poses a practical question and answers it with runnable code
• Original Python 2.7 and scikit-learn 0.16 notebooks preserved in an archive branch

Verdict

Newcomers who need a structured narrative from first dataset to production-style pipeline will get the most value here. If you already treat the scikit-learn docs as leisure reading, you will likely find the pace too gentle.