multickmeans: Consensus K-means clustering over mutiple K

In TankredO/ckmeans: consensus K-means clustering

Description

Consensus k means clustering for multiple ks

Usage

multickmeans(
  x,
  ks,
  n_rep = 50,
  p_pred = 1,
  p_samp = 1,
  save_kms = TRUE,
  hclust_options = list(),
  calc_bic = TRUE,
  ...
)

Arguments

x	matrix with samples in rows and features in columns
ks	vector of cluster numbers
n_rep	number of individual k means runs
p_pred	proportion of predictors used in every k means run
p_samp	proportion of samples used in every k means run
save_kms	logical, (or 'minimize') determining whether the k means object should be saved. This can be very memory demanding, depending on n_rep and x. If 'minimize', the kmeans objects are saved without column names, saving memory.
hclust_options	list of option passed to hclust, which is used to generate consensus clusters
calc_bic	logical, determining whether the BIC (Bayesian Information Criterion) should be calculated for the k means runs
...	arguments passed to kmeans

Details

Runs several independent k means clustering steps for several K, and combines information from the different runs to calculate consensus clusters using hierarchical clustering. The hierarchical clustering is based on the proportion of runs in which each pair of samples is placed in the same cluster, interpreted as distance.

Value

multickmeans object. ckms contains the list of ckmeans objects, ks is the vector of ks, bics is the vector of BICs.

Author(s)

Tankred Ott

TankredO/ckmeans documentation built on April 5, 2020, 12:59 a.m.