OPTICS

Ordering Points to Identify the Clustering Structure(OPTICS) Is Density-based Method for Clustering which is good at handling data with varied densities. It identifies high density samples and extends clusters from them, However unlike DBSCAN, it can detect meaningful clusters in data of varying density.

Optics uses two parameters:

  • Distance ()
  • Reachability Distance

Advantages:

  • Doesn't require the number of clusters to be set.
  • It's good at handling noisy data and Outliers.
  • It has no assumptions on shape and size of data.
  • It can identify clusters with varying densities.
  • It doesn't require distance radius() to be set, unlike DBSCAN.

Disadvantages:

  • It's not good at handling high dimensional data.
  • It's more computationally expensive than DBSCAN.
  • It only produces cluster ordering.