DBSCAN (density-based spatial clustering of applications with noise)

Also written: DBSCAN

A clustering method that grows clusters through crowded neighborhoods, so clusters can have any shape. Isolated points are labeled noise (−1) instead of being forced into a cluster.