DBSCAN groups observations that connect through dense neighborhoods. It can leave isolated records outside every cluster, and it doesn't require you to choose a cluster count first. The radius and minimum neighborhood size control that decision. You will fit density clusters, inspect the noise label and compare membership without attaching meaning to arbitrary group numbers. A group labeled zero is still a group; noise has label minus one. Distance units matter, so inputs with unrelated scales usually need preparation. This lesson uses a deliberately separated two-dimensional fixture so you can inspect the neighborhood behavior before handling ambiguous business segments.