DBSCAN

DBSCAN finds groups where many dots are close together and marks lonely dots as noise.

What it does

DBSCAN finds groups where many dots are close together and marks lonely dots as noise.

Density: How crowded an area is • Noise: Data that doesn't fit

How it works

• Look around each dot to count nearby dots.
• Connect dots that are in crowded areas.
• Mark far-away dots as noise.

Uses and applications

• A map app can find busy areas where many people visited.
• A factory can notice odd sensor readings that do not fit any normal group.
• A photo app can find clusters of stars or lights in a picture.

In short

It finds irregularly shaped groups and ignores noisy data.

Related methods

• K-Means: Groups points around a set number of centres instead of by density.
• Isolation Forest: Focuses on the unusual points that DBSCAN marks as noise.

Back to Patterns, anomalies and recommendations · All topics

The interactive demonstration and the Python walkthrough for this topic need JavaScript.