Patterns › Clustering

Compute k-means elbow (k=1..10)

The elbow plot charts total within-cluster sum of squares for k = 1..10 in k-means; the bend in the curve is a heuristic choice of k.

What is Compute k-means elbow (k=1..10)?

WSS always decreases as k increases (more clusters = tighter fit), so the question is when adding another cluster stops paying off. The 'elbow' — where WSS bends from steep to flat — is the heuristic answer, but the visual cut is subjective.

Principled alternatives: silhouette score (silhouette tool in our app), gap statistic, BIC-based selection via twostep. The elbow is fast and intuitive but less defensible for a paper.

When should I use Compute k-means elbow (k=1..10)?

Initial k pick for k-means.
Confirming a chosen k is in the right neighbourhood.
Switch to twostep when you want a defensible BIC-based pick.

What data does it need?

≥ 2 numeric feature columns.

What does it report?

Line plot of WSS vs k + raw values for export.

How do I interpret the result?

Smooth curves with no clear elbow ⇒ the data has no natural cluster structure or you need a different similarity metric.