DBSCAN Clustering

Harvard EPS-210 | Interactive tutorial — Explore density-based clustering with automatic noise detection

Clustering Result

Core
Border
Noise

ε-Neighborhoods

Nε(p) = {q ∈ D | dist(p,q) ≤ ε}
Points within ε-radius of p

k-Distance Graph

Density Map

Data Points

Click on the clustering canvas to add points

DBSCAN Parameters

0.80
ε (epsilon)
4
minPts
Epsilon (ε) 0.80
MinPts 4

Algorithm

Statistics

Points
0
Clusters
0
Core
0
Noise
0
Hover over a point to inspect
Point Type: --
Neighbors: --
Cluster: --

Quick Examples

How it works: DBSCAN finds clusters based on density. Core points have ≥ minPts neighbors within ε. Border points are within ε of a core point. Points that are neither are noise. Unlike K-means, it can find arbitrarily shaped clusters and automatically detects outliers.