Grade 12 — Data scientists II (modelling and communicating)¶
Tools: scikit-learn, pandas, Streamlit, Git/GitHub (Django optional) · Units: 8 · Lessons: 24 (+ projects)
Grade 12 is about modelling and communicating: the supervised machine-learning workflow, regression and classification, evaluation metrics, overfitting, a fairness audit, feature engineering, data storytelling, and a community capstone with a real "client" (ward office, health post, school, NGO). Health examples use the synthetic table (real DHS data cannot be redistributed); every notebook ends by asking what could go wrong for real people. Django is an optional stretch for students who already know web basics.
| Unit | Title | Lessons | Track | Project |
|---|---|---|---|---|
| 1 | The machine-learning workflow | 3 | B (code) | Workflow report |
| 2 | Regression and classification | 3 | B (code) | Explain a prediction |
| 3 | Evaluation metrics | 3 | B (code) | Choose the threshold |
| 4 | Overfitting and generalising | 3 | B (code) | Find the sweet spot |
| 5 | Fairness and bias audit | 3 | B (code + discussion) | Audit memo |
| 6 | Feature engineering | 3 | B (code) | Feature challenge |
| 7 | Communicating with data | 3 | A/B | Story and memo |
| 8 | Capstone with a community client | 3 | B/A (project) | Community capstone |
Every chapter now has detailed explanations with worked examples (all Python in the chapters is run in CI by
tools/check_chapter_code.py), common mistakes, a unit summary, practice exercises (★/★★/★★★, answers in the
teacher guide) and exercises with real data from docs/datasets/free-datasets.md,
each with an offline fallback. Unit 8 ends with an end-of-school checklist; the book's "After Grade 12" part continues
from there.
Flagship project: Capstone with a community client — a public write-up, model card and (optional) dashboard
Each unit folder holds: chapter.md (student), teacher-guide.md, project.md, quiz.yml, datasets.yml
(and notebooks/ where code is used). Nepali chapters: see community/translations/STATUS.md.