Home▸AI & Machine Learning▸Silhouette Score and the Elbow Method

📊 Silhouette Score and the Elbow Method

K-means needs a number of clusters before it can run. A practical walkthrough of the elbow method and the silhouette score, the two standard tools for choosing that number without just guessing.

AI & Machine Learning3DAdvanced60 FPS

A live k-means clustering lab: run clustering on a 3D point cloud, sweep the number of clusters k, and watch the elbow (inertia) curve and the silhouette-score curve build in real time.

🔬 What It Demonstrates

Inertia always decreases as k grows, so the "elbow" — not the minimum — is the signal. Silhouette score instead peaks near a genuinely good k, and can even fall as k keeps rising, giving a second, independent check.

🎮 How to Use

Pick a dataset, drag the k slider or hit Sweep to watch both bar charts respond, and switch point colouring to silhouette value to see exactly which points are poorly matched to their cluster.

💡 Did You Know?

On the "Uniform noise" dataset there is no real cluster structure at all — both curves go nearly flat, which is itself a useful diagnostic: no k is a good k.

⚙ Under the hood

K-means needs a number of clusters before it can run. A practical walkthrough of the elbow method and the silhouette score, the two standard tools for choosing that number without just guessing.

machine learningclusteringk-meanselbow methodsilhouette scoredata analysisalgorithmsThree.js

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install