Clustering Methods by Hand: Hierarchical, k-Medoids, Silhouette & DBSCAN
Distance matrices, single vs complete linkage, dendrogram cuts, silhouette scores, elbow vs silhouette, a PAM swap and DBSCAN core/border/noise, every step computed cell by cell on 12 stores.
The first lesson is free — no account needed.
About this course
A clustering methods course that segments the stores of one fictional retail chain, Tallow, by hand in a spreadsheet, for data analysts, data science students and anyone who runs clustering in Python and wants to see each step. You scale two features to 0–1 and build a 12 × 12 Euclidean distance matrix, run agglomerative clustering with single and complete linkage, cut a dendrogram at the largest gap, compute the silhouette score of every store and the average, compare the elbow rule with the best silhouette when choosing k, price the medoids and test one PAM swap, label core, border and noise points with DBSCAN using COUNTIF on the distance matrix, and profile the final segments with a scatter chart coloured by cluster. The first lab is free.
Lessons
- Distance Matrices: Euclidean, Manhattan & Why Scaling Comes FirstFree
- Agglomerative Clustering: Single vs Complete Linkage, Step by Step Pro
- Reading a Dendrogram: Merge Heights, the Largest Gap & Cluster Labels Pro
- Silhouette Score per Store: a, b, s and the Average Pro
- Choosing k: Elbow Rule vs Silhouette Pro
- k-Medoids (PAM): Total Cost and One Swap Step Pro
- DBSCAN by Hand: eps, minPts and Core, Border & Noise Points Pro
- Cluster Profiling and a Scatter Chart Coloured by Cluster Pro