Skip to main content

Feature engineering

Building and selecting features, the step that often does more for performance than the choice of model.

Course Duration: 5h

What You'll Learn

  • Handle missing values without introducing bias
  • Encode categorical variables according to their cardinality
  • Create business features that carry genuinely new information
  • Detect and eliminate data leakage before it distorts everything
  • Select a subset of features and justify that choice

Prerequisites

  • Python and pandas
  • Notions of supervised learning

Course Modules

  1. Why features matter more than the algorithm
  2. Missing values: dropping, imputing, indicating
  3. Scaling: normalization and standardization
  4. Encoding: one-hot, ordinal, target
  5. Cyclical features, dates and durations
  6. Text features: bag of words and n-grams
  7. Features from aggregations and time windows
  8. Data leakage: the most frequent cases
  9. Feature selection: filters, wrappers, importance
  10. Reproducible pipelines with scikit-learn

Assessment and certificate

The course ends with a 40-question exam covering every module. On success, a certificate of completion is issued; its number is verifiable on the platform.

Free courses, by contrast, end with a 5-question quiz and a preview of the certificate, without certification.