K-Means Clustering — Magical Learning Lab

Group similar things into k colourful families! Choose k, watch centroids dance, and see how points find their best cluster. Toggle 2D/3D, switch datasets, and explore the Elbow Method.

1. Choose k

2. Initialize

3. Assign

4. Update

Play Controls

Blobs Two Moons Rings

k (clusters) 3

Step
—

Current SSE
—

Centroids
—

Points
—

Switch to 3D Magic

What are we minimizing?

K-Means pulls each point towards the nearest center. We measure goodness with SSE (sum of squared errors). Smaller is better!

Interactive Visualizer

2D Constellation

Unassigned C1 C2 C3 C4 C5

Elbow Method — choose a good k

Look for the “elbow”: the point where adding more clusters no longer improves SSE a lot.

Pick k — choose how many groups you expect (try 2–6).

Initialize — place k starting centroids (random).

Assign — each point joins the nearest centroid.

Update — move each centroid to the average of its points.

Repeat — Assign ↔ Update until centroids barely move.

Everyday Examples

Group students by study habits (time vs. scores)
Segment shoppers by spending & visits
Compress images by clustering colors
Bundle similar news articles