Discover clusters based on density. DBSCAN automatically finds clusters of arbitrary shapes and identifies outliers without needing to specify the number of clusters.
Speed:
ε:
MinPts:
Dataset
Algorithm
Parameters
ε (epsilon): neighborhood radius MinPts: minimum points for dense region
Point Types
• Core: ≥MinPts neighbors within ε • Border: <MinPts neighbors, reachable from core • Noise: not reachable from any core
Clustering
Connect density-reachable core points and their borders into clusters.