How it works...

DBSCAN works by treating datapoints as groups of dense clusters. If a point belongs to a cluster, then there should be a lot of other points that belong to the same cluster. One of the parameters that we can control is the maximum distance of this point from other points. This is called epsilon. No two points in a given cluster should be further away than epsilon. One of the main advantages of this method is that it can deal with outliers. If there are some points located alone in a low-density area, DBSCAN will detect these points as outliers as opposed to forcing them into a cluster.