Density-based clustering algorithms such as DBSCAN work by building up on the
definition of density as the number of points within a specified radius. DBSCAN,
for instance, defines three kinds of points: core points are those that have more than
a specified number of neighbors within a given distance; border points have fewer
than the specified number but belong to a core point neighborhood; and noise points
are those that are neither core or border. The algorithm iteratively removes noise
points and performs clustering on the remaining points.