Teach scikit-learn as a consistent machine learning workflow rather than a list of commands to memorise. This Unit 06 pack transforms Chapter 6 of AI & Machine Learning: Foundations & Data – Volume 1 into four structured, classroom-ready lessons covering the Estimator API, model families, safe hyperparameter tuning, pipelines, custom transformers, model persistence and imbalanced data.

Students learn how fit, transform, predict and score represent different responsibilities, how predictors, transformers, pipelines and meta-estimators fit together, and why model selection must be supported by evidence. They compare Ridge, Lasso and Random Forest, use GridSearchCV and RandomizedSearchCV safely, keep preprocessing inside pipelines and protect the final test set.

The unit also introduces responsible model reuse through joblib persistence, feature schema, package versions, retesting and class imbalance. Students compare recall, precision and F1, examine class weighting and SMOTE, and meet optional advanced tools such as IterativeImputer and RFE.

Designed as the classroom teaching companion to AI & Machine Learning: Foundations & Data – Volume 1, the package includes a Teacher Guide, four-lesson plan, Teacher Handbook, Student Notes, Worksheet, Answer Key, Activity Cards, Exit Tickets, an editable 34-slide deck and a Read Me First guide. It supports no-code reasoning and light Python code reading.

The no-prep structure supports teaching, group work, assessment and revision. Trustworthy scikit-learn work depends on method responsibility, allowed data, validation evidence, saved assumptions and clear limitations.

By the end of this unit, students will be able to:

• Explain the scikit-learn Estimator API pattern
• Distinguish predictors, transformers, pipelines and meta-estimators
• Compare model families using evidence and limitations
• Explain Ridge, Lasso and Random Forest
• Use GridSearchCV and RandomizedSearchCV reasoning safely
• Explain why preprocessing belongs inside a pipeline
• Identify leakage and test-set misuse
• Explain custom transformers and reusable preprocessing
• Describe responsible model persistence with joblib
• Recognise version and feature-schema risks
• Evaluate imbalanced classification using recall, precision and F1
• Explain class weighting and safe SMOTE use
• Interpret IterativeImputer and RFE as optional advanced tools
• Write a cautious model claim with evidence and limitations

Lesson 1 – The Estimator API
Understand fit, transform, predict and score, and identify object roles.

Lesson 2 – Datasets, Linear Models and Ensembles
Compare dataset sources and choose model families using evidence.

Lesson 3 – Hyperparameter Tuning Without Leakage
Use grid search, random search and pipelines while protecting validation and test evidence.

Lesson 4 – Custom Transformers, Persistence and Imbalance
Connect reusable preprocessing, saved workflows, versioning and minority-class metrics to responsible deployment.