Description

Build a complete, responsible machine-learning classification workflow with Python and scikit-learn.

This hands-on practical guides you from defining a prediction task through model comparison, holdout evaluation, threshold selection, subgroup checks, and evidence-aware reporting. You will work with cleaned synthetic service-delivery data and compare a prevalence-only baseline, logistic regression, and a shallow decision tree.

What you will learn

Distinguish prediction from explanation and causal inference.
Define targets, features, observation units, and prediction moments.
Identify identifier, proxy, temporal, and preprocessing leakage.
Create reproducible stratified training and test sets.
Build mixed-type preprocessing with scikit-learn pipelines.
Compare models using cross-validation and an untouched holdout set.
Interpret accuracy, precision, recall, F1, ROC AUC, confusion matrices, predicted probabilities, and decision thresholds.
Check subgroup performance and communicate responsible-use boundaries.

What you will produce

A feature and leakage review.
Cross-validation and holdout performance tables.
Confusion-matrix, ROC, precision-recall, and threshold figures.
A subgroup performance table and coefficient table.
A saved end-to-end model pipeline.
A model card, decision log, and portfolio-ready project statement.

Tools: Python, pandas, NumPy, scikit-learn, Matplotlib, seaborn, joblib, and Quarto.

Important: This practical is for education, portfolio evidence, and local technical testing. Its synthetic data and demonstration outputs are not validated for operational, clinical, or other real-world decisions affecting people.