Feature Engineering and Model Evaluation

Curriculum

1. 01Features and Point-in-Time ThinkingWhy good features matter more than clever algorithms, how to frame a prediction around a snapshot date, and how to define a target like churn precisely enough to build on.25 min
2. 02Features from EventsTurn a stream of transactions into one row per customer with recency, frequency, value, time windows, trends and failure rates, and check each feature actually relates to churn.15 min
3. 03Leakage Across TimeSee how a random train/test split and one careless feature make a model look far better than it is, and learn the time-based split that gives an honest answer.30 min
4. 04Gradient Boosting and Model ComparisonTrain gradient-boosted trees, the workhorse of tabular machine learning, compare them fairly with logistic regression on a time-based test, and decide when the extra complexity earns its keep.20 min
5. 05CalibrationCheck whether a model's probabilities mean what they say, measure calibration with a reliability table and the Brier score, and understand why it matters whenever probabilities drive money.25 min
6. 06Lift and TargetingEvaluate a model the way the business will use it, with gains and lift for the top of the list, and choose how many customers to contact from the cost of a call and the value of a save.25 min
7. 07Drift and RetrainingDetect when the world a model learned from has changed, find which customers changed, and retrain on recent snapshots so the model keeps up.25 min
8. 08Final ProjectPlan your final project, an end-to-end churn model with point-in-time features, time-based validation, calibration, targeting and monitoring, and start with a feature of your own.20 min
9. Final assessmentPass it (60% or more) to earn the Feature Engineering and Model Evaluation course completion badge.