covsurf: Combining ClustOfVar and VSURF

In robingenuer/CoVVSURF: Combines clustering of variables with feature selection using random forests

Description

This function selects groups of informative input variables to predict a response variable.

Usage

covsurf(X, y, kval = 2:ncol(X), tree = NULL, nse = 1, ncores = 1, ...)

Arguments

X	dataframe of input variables
y	vector of responses
kval	vector of number of classes to try
tree	optional tree given by hclustvar
nse	number of standard-deviation to add to select minimum of OOB rate
ncores	number of cores to use for parallel computing
...	passed to VSURF

Value

kopt	the optimal number of groups of variables
ptree	the partition in kopt clusters of the dendrogram of CoV.
vsurf_ptree	VSURF applied to the kopt synthetic variables.
vsel	synthetic variables selected by VSURF.
csel	groups of variables selected by VSURF.
rfsel	RF applied to the selected synthetic variables
rfclust	RF applied to all the synthetic variables.
oob	a matrix with mean OOB error (first column) and OOB standard deviation (second column).

References

Combining clustering of variables and feature selection using random forests: the CoV/VSURF procedure, Marie Chavent, Robin Genuer, Jerome Saracco, hal-01345840

Examples

data(don60)
kval <- c(2:15, seq(from = 20, to = ncol(X), by = 10))
don60covs <- covsurf(X, y, kval)
plot(don60covs)

robingenuer/CoVVSURF documentation built on May 27, 2019, 11:38 a.m.