build.clusters: Apply cluster analysis to build variable groups

In StephanSeifert/SurrogateMinimalDepth: Surrogate minimal depth variable importance

Apply cluster analysis to build variable groups

Description

This function generates variables groups of relation information that was obtained by var.relations function applying getLinkCommunities.

Usage

build.clusters(rel, hcmethod = "ward.D")

Arguments

rel	a list containing variables, surr.res, threshold, and var. This is the output of var.relations function.
hcmethod	the hierarchical clustering method that is used. (see getLinkCommunities)

Value

a data frame containing the variable names and their associated clusters.

Examples

# read data
data("SMD_example_data")


 # get trees and variable names
 x = SMD_example_data[,2:ncol(SMD_example_data)]
 y = SMD_example_data[,1]
 allvariables = colnames(x)# extract variables names
 nvar = length(allvariables)   # count number of variables
 set.seed(42)
 RF = ranger::ranger(data = SMD_example_data, dependent.variable.name = "y", num.trees = 10, keep.inbag = TRUE,mtry = floor(nvar^(3/4)), min.node.size = 1)
 trees = getTreeranger(RF = RF, ntree = 10)
 trees.lay = addLayer(trees)
 trees.surr = addSurrogates(RF = RF, trees = trees.lay, s = 10, Xdata = x, num.threads = NULL)

 # investigate variable relations
 rel=var.relations(forest = list(trees = trees.surr, allvariables = allvariables), variables = allvariables, candidates = allvariables, t = 10)
 groups = build.clusters(rel)

StephanSeifert/SurrogateMinimalDepth documentation built on Aug. 7, 2023, 1:59 a.m.