Phase 3 · Machine learning · Foundations
Feature engineering
Turn raw data into signals a model can use.
Open this lesson Create a free account
Why it matters
A model can only learn from what it's shown. Turning raw columns into signals that match how the world works (a rate instead of two counts, "days since" instead of a date, a log instead of a wildly skewed number) is often worth more than switching to a fancier algorithm. Marketers already think in engineered features: RFM scores, engagement rates, cohorts. This lesson turns that instinct into code and shows how to check that a feature earns its place.
What you'll be able to do
- Create features from ratios, flags, logs and combinations of columns
- Build recency, frequency and monetary (RFM) features for customers
- Measure whether a new feature actually helps a model
- Know where hand-made features matter less (deep learning) and where they still dominate
Before this lesson
Next in this phase: Train, validation and test sets