Intro
Silhouette Score evaluates cluster quality by measuring cohesion (how close points are to others in their cluster) and separation (how far points are from other clusters).
Silhouette = (separation - cohesion) / max(separation, cohesion)
| Score Range | Interpretation |
|---|---|
| 0.50 – 1.00 | Clear, well-separated clusters |
| 0.26 – 0.49 | Some clustering structure with overlap |
| 0.00 – 0.25 | Weak clustering or heavy overlap |
| < 0 | Points may be assigned to the wrong cluster |