filteredFit: A generic fit method with a filtered step for feature...
In FRESA.CAD: Feature Selection Algorithms for Computer Aided Diagnosis

filteredFit

R Documentation

A generic fit method with a filtered step for feature selection

Description

Sequential application of decorrelation, scaling, feature selection, and PCA/Whitening then fit

Usage

	
    filteredFit(formula = formula, 
	            data=NULL,
	            filtermethod=univariate_KS,
	            fitmethod=e1071::svm,
	            filtermethod.control=list(pvalue=0.10,limit=0),
	            Scale="none",
				PCA=FALSE,
				WHITE=c("none","CCA"),
	            DECOR=FALSE,
	            DECOR.control=list(thr=0.80,method="fast",type="NZLM"),
	            ...
	            )

Arguments

`formula`	the base formula to extract the outcome
`data`	the data to be used for training the KNN method
`filtermethod`	the method for feature selection
`fitmethod`	the fit function to be used
`filtermethod.control`	the set of parameters required by the feature selection function
`Scale`	Scale the data using the provided method
`PCA`	Decorrelate the input data using PCA
`WHITE`	Whittening process: "PCA" or "CCA"
`DECOR`	Decorrelate the input data estimating the UPSTM
`DECOR.control`	Parameters to the decorrelation function
`...`	parameters for the fitting function

Value

`fit`	The fitted model
`filter`	The output of the feature selection function
`selectedfeatures`	The character vector with all the selected features
`usedFeatures`	The set of features used for training
`parameters`	The parameters passed to the fitting method
`asFactor`	Indicates if the fitting was to a factor
`classLen`	The number of possible outcomes