Data Science: Building Machine Learning Models

Rafael Irizarry teaches it first-principles: you build a working movie-recommendation system while learning why cross-validation, regularization and PCA matter, rather than just calling a library. That statistical framing is the value, and a useful contrast with code-first Python courses.

It is introductory but assumes the R and data-analysis grounding from earlier courses in the series, so it is not a standalone first course, and if you prefer Python or want a fast applied tutorial it is the wrong fit. It is free to audit, with an optional paid HarvardX certificate. Harvard maintains the series (as of 2026).

Data Science: Building Machine Learning Models is part of HarvardX's renowned Professional Certificate in Data Science. Over about eight self-paced weeks you learn the fundamentals of machine learning — popular algorithms, training and test data, cross-validation to avoid overtraining, regularization, and principal component analysis — by building a working movie recommendation system from scratch.