feature_selection: Perform feature selection

In specmine: Metabolomics and Spectral Data Analysis and Mining

Description

Perform feature selection on the dataset.

Usage

feature_selection(dataset, column.class, method = "rfe", 
functions, validation = "cv", repeats = 5, number = 10, 
subsets = 2^(2:4))

Arguments

dataset	list representing the dataset from a metabolomics experiment.
column.class	string or index indicating what metadata to use.
method	method used for feature selection. Possible values are "rfe" (recursive feature elimination) and "filter" (Selection by filter - sbf) from caret's package.
functions	a list of functions for model fitting, prediction and variable importance/filtering.
validation	the external resampling method: boot, cv, LOOCV or LGOCV (for repeated training/test splits.
repeats	for repeated k-fold cross-validation only: the number of complete sets of folds to compute.
number	either the number of folds or number of resampling iterations.
subsets	a numeric vector of integers corresponding to the number of features that should be retained (rfe only).

Value

caret's result from rfe or sbf.

Examples

  ## Example of feature selection using rfe and sbf
  library(caret)
  library(specmine.datasets)
  data(cachexia)
  rfe.result = feature_selection(cachexia, "Muscle.loss", 
               method="rfe", functions = caret::rfFuncs, 
               validation = "cv", number = 3, 
               subsets = 2^(1:6))
  sbf.result = feature_selection(cachexia, "Muscle.loss", 
	       method="filter", functions = caret::rfSBF, 
	       validation = "cv")

specmine documentation built on Sept. 21, 2021, 5:06 p.m.