Kmeans.sim.noise: Function to compute similarity indices using noise injection...

In mosclust: Model Order Selection for Clustering

Function to compute similarity indices using noise injection techniques and kmeans clustering.

Description

A vector of similarity measures between pairs of clusterings perturbed with random noise is computed for a given number of clusters. The variance of the added gaussian noise, estimated from the data as the perc percentile of the standard deviations of the input variables, the percentile itself and the similarity measure can be selected.

Usage

Kmeans.sim.noise(X, c = 2, nnoisy = 100, perc = 0.5, s = sFM, 
                 distance = "euclidean", hmethod = NULL)

Arguments

X	matrix of data (variables are rows, examples columns)
c	number of clusters
nnoisy	number of pairs of noisy data
perc	percentile of the standard deviations to be used for the added gaussian noise (def. 0.5)
s	similarity function to be used. It may be one of the following: - sFM (Fowlkes and Mallows) - sJaccard (Jaccard) - sM (matching coefficient) (default Fowlkes and Mallows)
distance	actually only the euclidean distance is available "euclidean" (default)
hmethod	parameter used for internal compatibility.

Value

vector of the computed similarity measures (length equal to nnoisy)

Author(s)

Giorgio Valentini [e-mail]

See Also

Kmeans.sim.projection, Kmeans.sim.resampling, perturb.by.noise

Examples

library("clusterv")
# Synthetic data set generation
M <- generate.sample6 (n=20, m=10, dim=600, d=3, s=0.2);
# computing a vector of similarity indices with 2 clusters:
v2 <- Kmeans.sim.noise(M, c = 2, nnoisy = 20,  s = sFM)
# computing a vector of similarity indices with 3 clusters:
v3 <- Kmeans.sim.noise(M, c = 3, nnoisy = 20,  s = sFM)
# computing a vector of similarity indices with 2 clusters using the Jaccard index
v2J <- Kmeans.sim.noise(M, c = 2, nnoisy = 20,  s = sJaccard)
# 2 clusters using 0.95 percentile (more noise)
v095 <- Kmeans.sim.noise(M, c = 2, nnoisy = 20,  s = sFM, perc=0.95)

mosclust documentation built on June 8, 2025, 11:23 a.m.