📄️ Feature engineering
Building and selecting features, the step that often does more for performance than the choice of model. 5 hours, applied projects, a 40-question exam and a verifiable certificate.
📄️ 1. Features or algorithm
Module 1 of the Feature Engineering premium course: why feature quality caps performance, what a good feature contributes, and the real place of this step in a project.
📄️ 2. Missing values
Module 2 of the Feature Engineering premium course: why a value is missing, the MCAR, MAR and MNAR mechanisms, imputation strategies and the missingness indicator.
📄️ 3. Scaling
Module 3 of the Feature Engineering premium course: which models require scaling, differences between standardization, min-max normalization and robust scaling, and distribution transformations.
📄️ 4. Categorical encoding
Module 4 of the Feature Engineering premium course: why naive encoding induces a false order, one-hot and its cardinality limit, ordinal, target and frequency encoding.
📄️ 5. Dates and cycles
Module 5 of the Feature Engineering premium course: decomposing a date into useful features, encoding cycles with sine and cosine, and building durations and recency.
📄️ 6. Text features
Module 6 of the Feature Engineering premium course: normalizing text, bag of words, TF-IDF weighting, n-grams and the simple descriptive features often forgotten.
📄️ 7. Aggregations and windows
Module 7 of the Feature Engineering premium course: aggregating by group, building deviation-from-habit features, and computing rolling windows and lags without leakage.
📄️ 8. Data leakage
Module 8 of the Feature Engineering premium course: the four families of leakage, warning signals, the investigation method and structural remedies.
📄️ 9. Feature selection
Module 9 of the Feature Engineering premium course: why reduce the number of features, filter, wrapper and embedded methods, and permutation importance.
📄️ 10. Pipelines
Module 10 of the Feature Engineering premium course: assembling preprocessing and model in a Pipeline, handling heterogeneous columns with ColumnTransformer, and industrializing a custom transformer.
📄️ Recap and exam
Complete recap of the Feature Engineering premium course: handling, encoding, building, cleaning and selecting features module by module, then the 40-question exam.