SigmaAnalytics & Tech
G3PythonAdvancedAbout 180 min to complete

Clustering

Every model so far learned from labelled examples — you told it the right answer (species, survived, churned) and it learned to reproduce it. Clustering is different: it is unsupervised.

Notebook

About 180 minutes to complete

Enrol to read the notebook and download it to run.

What this module covers

Every model so far learned from labelled examples — you told it the right answer (species, survived, churned) and it learned to reproduce it. Clustering is different: it is unsupervised.

By the end of it

  • Explain unsupervised learning and how clustering differs from classification
  • Run k-means, interpret its centroids and inertia, and choose k with the elbow and silhouette methods
  • Understand why scaling is essential for distance-based clustering
  • Run and read hierarchical (agglomerative) clustering with a dendrogram
  • Run DBSCAN to find arbitrarily-shaped clusters and label noise/outliers
  • Evaluate clusterings with the silhouette score (no labels) and adjusted Rand index (when labels exist)
  • Use PCA to reduce dimensions for 2-D cluster visualisation