INFO
Determines the optimal number of clusters by plotting within-cluster variance as a function of the number of clusters.
How It Works
- Plot the sum of squared errors (SSE) for different values of
- Look for the “elbow” point where SSE starts to level off
- That point suggests the optimal number of clusters
What to Look For
- Visual method, not a strict formula
- Works best with convex clusters
- Often used with K-Means