Parameters for Feature Selection
SelectParams-class.Rd
Collects and checks necessary parameters required for feature selection. Either one function is specified or a list of functions to perform ensemble feature selection. The empty constructor is provided for convenience.
Constructor
SelectParams(featureRanking, characteristics = DataFrame(), nFeatures = 20, minPresence = 1, intermediate = character(0), subsetToSelections = TRUE, tuneParams = list(nFeatures = seq(10, 100, 10)), ...)
Creates a
SelectParams
object which stores the function(s) which will do the selection and parameters that the function will use.featureRanking
A character keyword referring to a registered feature ranking function. See
available
for valid keywords.characteristics
A
DataFrame
describing the characteristics of feature selection to be done. First column must be named"charateristic"
and second column must be named"value"
. If using wrapper functions for feature selection in this package, the feature selection name will automatically be generated and therefore it is not necessary to specify it.nFeatures
Default:
20
. The number of top-ranked features to choose. Can also beNULL
if a vector of top numbers is specified totuneParams
for the list element namednFeatures
.minPresence
Default:
1
. If a list of functions was provided, how many of those must a feature have been selected by to be used in classification. 1 is equivalent to a set union and a number the same length asfeatureSelection
is equivalent to set intersection.intermediate
Character vector. Names of any variables created in prior stages by
runTest
that need to be passed to a feature selection function.subsetToSelections
Whether to subset the data table(s), after feature selection has been done.
tuneParams
A list specifying tuning parameters to try during feature selection. A list element named
nFeatures
is used to represent a variety of top-n ranked features to try. Other names of the list are the names of the parameters of the ranking function and the vectors are the values of the ranking function's parameters to try. All possible combinations are generated....
Other named parameters which will be used by the selection function. If
featureSelection
was a list of functions, this must be a list of lists, as long asfeatureSelection
.
Summary
selectParams
is aSelectParams
object.show(SelectParams)
: Prints a short summary of whatselectParams
contains.
Examples
#if(require(sparsediscrim))
#{
SelectParams("KS")
#> An object of class 'SelectParams'.
#> Selection Name: Kolmogorov-Smirnov Test.
# Ensemble feature selection.
SelectParams(list("Bartlett", "Levene"))
#> An object of class 'SelectParams'.
#> Ensemble Selection: Bartlett Test, Levene Test.
#> Minimum Functions Selected By: 1
#}