bootclus: Generating a credal partition by bootstraping Gaussian...
In evclust: Evidential Clustering

bootclus

R Documentation

Generating a credal partition by bootstraping Gaussian Mixture Models

Description

bootclus generates a credal partition by bootstrapping Gaussian Mixture Models.

Usage

bootclus(
  x,
  conf = 0.9,
  B = 500,
  param = list(G = NULL),
  type = "pairs",
  Omega = FALSE
)

Arguments

`x`	attribute matrix or data frame of size (n,p).
`conf`	confidence level (default: 0.90).
`B`	number of bootstrap samples (default=500)
`param`	list of arguments passed to function `Mclust` in addition to 'data'.
`type`	Type of focal sets ("simple": `\emptyset`, singletons and `\Omega`; "full": all `2^c` subsets of `\Omega`; "pairs": `\emptyset`, singletons, `\Omega`, and all or selected pairs). Argument passed to `makeF`.
`Omega`	Logical. If TRUE, `\Omega` is a focal set. Default is FALSE. Argument passed to `makeF`.

Details

This function uses the mclust package to generate and bootstrap the mixture models.

Value

A list with the following components:

clus: An object of class 'Mclust' returned by Mclust.
Clus: An object of class 'credpart' providing the output credal partition.
CI: An array of dimension (2,n,n) containing the confidence intervals on pairwise probabilities.
BelPl: An array of dimension (2,n,n) containing the pairwise Bel-Pl intervals.
Time: A matrix of size (3,5) containing the computing time as returned by function proctime for (1) the parameter estimation and bootstrap, (2) the computation fo the quantiles on pairwise probabilities, and (3) the computation of the credal partition.

References

T. Denoeux. Calibrated model-based evidential clustering using bootstrapping. Information Sciences, Vol. 528, pages 17-45, 2020.

Examples

## Example with the Faithful geyser data
## Not run: 
data("faithful")
X<-faithful
param=list(G=3)
res.faithful<-bootclus(X,conf=0.90,B=100,param=param)
## Plot the results
plot(res.faithful$Clus,X)

## End(Not run)

evclust documentation built on Nov. 9, 2023, 5:05 p.m.