module

Supervised Learning by Building

OBJECTIVE: learn every major supervised algorithm by building it on both projects — classification (German Credit) and regression (Electricity Demand) — and finish with a tuned, explained, deployed model. WHAT WE COVER: • How training works: parameters, losses, optimisation, learning curves, the scikit-learn API • An evaluation harness: metrics, ROC/PR, thresholds and costs, calibration, cross-validation • Choosing an algorithm; linear models and regularisation; KNN, Naive Bayes, LDA/QDA, SVM • Decision trees, Random Forest, Extra Trees, boosting (XGBoost, LightGBM, CatBoost), MLP (glimpse) • Ensembles and hybrids; model comparison, error analysis, SHAP and fairness • Tuning with grid/random search and Optuna; final one-time test • Saving, pipelines, a Gradio demo and a capstone AFTER THIS MODULE YOU CAN: • Choose, train, tune and compare supervised models honestly • Explain a model's decisions and check its fairness • Package a model with a demo