DBSCAN
DBSCAN clusters by density: it separates high-density regions into clusters, discovers arbitrarily shaped clusters, labels noise points, and needs no preset cluster count.
Purpose
To cluster by density, discovering arbitrarily shaped clusters and outliers.
Mechanism
Core point: at least minPts points within ε radius. Border point: within a core point's neighborhood but not itself core. Noise: neither core nor border. Points density-reachable from a core point form one cluster.
Use cases
- Spatial data clustering
- Outlier detection
- Arbitrarily shaped clusters