Feature Engineering for Machine Learning

"Feature Engineering for Machine Learning" by Soledad Galli (Udemy) is the most complete dedicated course on tabular feature engineering in Python, and it is worth taking if you are an intermediate data scientist or ML engineer who wants to systematize preprocessing - but it is the wrong first course because it assumes you already know regression and tree models. It is a focused, code-heavy course built around imputation, categorical encoding, variable transformation, discretization, outlier handling and date/datetime features, implemented with pandas, scikit-learn and the instructor's own open-source Feature-engine library. It holds a strong 4.6/5 rating on Udemy (roughly 3,500-3,700 ratings reported across snapshots; the platform/Class Central listing shows 4.6) and is one of the most established dedicated feature-engineering courses online, taught by the author of the Packt "Python Feature Engineering Cookbook" and a two-time LinkedIn Top Voice in Data Science. The most common complaints are that some coding walkthroughs run long and the prior knowledge it assumes (linear/logistic regression, trees) makes it a poor fit for true beginners. It is best treated as a practical reference for preprocessing skills rather than an introduction to machine learning itself.

It is an excellent, well-reviewed pick for its narrow target audience (intermediate practitioners who want to systematically improve data preprocessing), but it deliberately skips ML fundamentals and assumes you already know common predictive models, so it is the wrong starting point for beginners.

Best for: Working or aspiring data scientists, ML engineers and analysts who already understand basic machine learning (linear/logistic regression, decision trees, random forests) and want a comprehensive, hands-on catalog of feature engineering techniques they can apply to real tabular datasets and Kaggle-style competitions. Especially useful for people who want to learn or adopt the Feature-engine library and build reusable scikit-learn preprocessing pipelines.

Skip if: Complete beginners to machine learning or Python, people who want deep learning / NLP / computer vision feature work (this is tabular-data focused), and learners who prefer short, high-level conceptual overviews over long, detailed code walkthroughs. It is also redundant for senior practitioners who already have a mature preprocessing workflow.