1

I need to cluster (or group) large one dimensional data sets into a set of fixed bins. I started out using K-means, but I want to look into other approaches. Two that I have found are Mean Shift and KDE, but when I actually looked into the details, they were both Parzen window non parametric estimators.

Is there a difference between the two?

Also does DBSCAN do the same thing as well?

Has QUIT--Anony-Mousse
  • 39,639
  • 7
  • 61
  • 96
Skander H.
  • 10,602
  • 2
  • 33
  • 81

1 Answers1

1

Mean shift essentially tries to appproximate the local maxima of the KDE using a heuristic search.

Full KDE can do much more than that.

DBSCAN looks for connected components. It uses a binary density model (i.e. dense enough, or noise).

Has QUIT--Anony-Mousse
  • 39,639
  • 7
  • 61
  • 96