Part 5 — Grade 12: Data scientists II¶
Modelling, evaluation, fairness, communication — and a real client.
Grade 12 brings everything together. You will follow a professional machine-learning workflow, choose between regression and classification, evaluate models by the mistakes they make, avoid overfitting, audit models for fairness, engineer features without leakage, communicate to decision-makers, and complete a capstone that serves a real client responsibly.
| Unit | Title | Lessons | Project |
|---|---|---|---|
| 1 | The machine-learning workflow | 3 | Workflow report |
| 2 | Regression and classification | 3 | Explain a prediction |
| 3 | Evaluation metrics | 3 | Choose the threshold |
| 4 | Overfitting | 3 | Find the sweet spot |
| 5 | Fairness and a bias audit | 3 | Audit memo |
| 6 | Feature engineering | 3 | Feature challenge |
| 7 | Communicating with data | 3 | Story and memo |
| 8 | Capstone with a client | 3 | flagship capstone |
Tracks: Track B (scikit-learn); Unit 7 and the capstone's scoping, ethics and communication work for every student.
Pacing: 24 lessons + projects ≈ 40 periods, with the capstone running over several weeks.
Key idea of the year: a model is only as good as the decision it supports — judge it against current practice, for every group, and say honestly when a simple rule is better.