Unit ML10-U05 · Predictive machine learning and responsible validation
Imbalance, calibration, and explanation
Analyze decision errors across relevant slices.
Outline only
This is an outline, not a lesson yet
The full lesson for this unit has not been written. This page shows what it is planned to teach, how long it is planned to take, and the reading chosen to support it. No explanation, worked example, quiz or task is published for it yet. See content status.
Planned outcome
Analyze decision errors across relevant slices.
How it would be assessed
Assessed outcome: Analyze decision errors across relevant slices.
Evidence: A task artifact demonstrating the stated action, a verification or comparison record, and a short explanation of uncertainty or failure. Use a new input for the completion check. Allocate the scheduled time across reading, demonstration, lab, quiz, and feedback rather than treating it as reading alone.
Planned time and place
- Planned time
- 5 h in total, including practice and assessment (a planning figure, not tested with learners)
- Module
- ML10: Predictive machine learning and responsible validation
- Chapter
- ML10-C03: Decision errors and final validation
- Prerequisite modules
- MA10, D30
- Used in programs
- P07
Selected reading for this topic
Core reading
- scikit-learn User Guide Model selection, preprocessing, metrics, pipelines, and task-appropriate estimators.Try it: Analyze decision errors across relevant slices.About this reference (W20)
Supporting reading
- An Introduction to Statistical Learning, with Applications in Python Statistical learning, linear methods, resampling, trees, and relevant Python labs.Try it: Analyze decision errors across relevant slices.About this reference (B05)
- Model Cards for Model Reporting Document intended use, performance slices, evaluation conditions, and limitations.Try it: Analyze decision errors across relevant slices.About this reference (P05)
Reading is chosen to support the outcome; none of it is compulsory homework. Links go to the owners' own sites and some resources cost money. Reviewed 2026-11-17.