Session Ready
Exercise

DBSCAN

DBSCAN is another very popular clustering algorithm, belonging to density-based algorithms.

For beginners it can seem very attractive because it doesn't require the number of clusters to be defined in advance.

But there's no free lunch and relying on DBSCAN to find the right number of clusters completely on its own can be a big trap.

Let's illustrate this by playing with DBSCAN's hyper-parameter eps, which defines the maximum distance between points within the same cluster.

Instructions 1/3
undefined XP
  • 1
  • 2
  • 3
  • Set the value of eps to 0.1 and check the clustering result.