What is it?
DBSCAN, which stands for Density-based Spatial Clustering of Application with Noise, is a density-based Clustering algorithm, which determine clusters as points which are close together, hence the density-based approach.
DBSCAN makes the assumption that high-density regions have lower-density regions separating them from other high-density regions, which create a sparse area around all clusters, that can be considered noise data or even outliers.
How does it work?
Differently from K-Means, DBSCAN makes short work of uneven clusters and manifolds, and can handle large datasets with moderate number of clusters, and it works well with noisy data and outliers, but not as much as BIRCH Clustering.
