Unsupervised Learning by Building (one technique, two mini projects)

OBJECTIVE: learn unsupervised learning one technique at a time, each with its own pair of mini projects. WHAT WE COVER: • Foundations: learning without labels and preparing data for distances • Clustering: k-means, hierarchical, DBSCAN/HDBSCAN, Gaussian mixtures (customer RFM segments, electricity day types) • Dimensionality reduction: PCA, t-SNE, UMAP (customer map, handwritten digits map) • Anomaly detection: statistical rules, Isolation Forest, LOF, One-Class SVM (credit-card fraud, load anomalies) • Association rules: Apriori / FP-Growth (market basket analysis) • Fair evaluation, optional industry projects and a capstone AFTER THIS MODULE YOU CAN: • Pick the right unsupervised technique for a business question • Build, evaluate and explain segments, maps, anomalies and rules