refine: Refine a clustering solution using the CluReAL algorithm

In unmnn/CluReAL: Clustering Refinement ALgorithm

Description

Run the CluReAL V2 algorithm on a data matrix and a vector of cluster membership indices. The output is a list with the refined cluster membership indices and the the cluster context.

Usage

refine(
  x,
  y,
  cc,
  gv,
  rc,
  rep = 0,
  min_rdens = -0.8,
  min_mass = 0.005,
  out_sens = 0.75,
  prun_level = 0
)

Arguments

x	The input data (numeric matrix).
y	Cluster membership indices (integer vector). -1 indicates noise.
cc	The cluster context.
gv	The GOI cluster validity indices object.
rc	The refinement context.
rep	The number of refinement repetitions.
min_rdens	The minimum density threshold (relative to the overall density) of a cluster to not be considered noise.
min_mass	The relative minimum size of a cluster to not be considered noise.
out_sens	Outlier sensitivity. Specifies how far away outliers must be from the centroids to avoid trying to merge them with the clusters.
prun_level	Pruning level. 0: normal (default), 1: high cluster overlap expected, 2: very high cluster overlap expected

Details

For a description of the algorithm, see:

Iglesias, Felix, Tanja Zseby, and Arthur Zimek. "Clustering refinement." International Journal of Data Science and Analytics (2021): 1-21. URL: https://doi.org/10.1007/s41060-021-00275-z.

Value

A list with two elements:

• y: Vector of the refined cluster membership indices.

• cc: Refined cluster context

Examples

# TODO

unmnn/CluReAL documentation built on Dec. 23, 2021, 2:01 p.m.