cvSCAD: Crossvalidation for SCAD SVM classification and feature...

In bootfs: Derive Robust Feature Sets for Two or Multiclass Classification Problems

Description

Main wrapper for call of SCAD SVM classification.

Usage

cvSCAD(logX, groupings, DIR, params = 
		list(seed = 123, ncv = 5, 
			repeats = 10, maxiter = 1000, 
			maxevals = 500, fs.method="scad"))

Arguments

logX	The data matrix. Samples in rows, features in columns.
groupings	A list. Each list element is a named vector (length equals the number of samples), holding group assignments for each sample (either 1 for group A and -1 for group B).
DIR	The output base directory.
params	A parameter list with the following elements. seed: a random seed. ncv: the number of folds in the crossvalidation. repeats: number of repeats of the CV. maxiter: SCAD parameter for maximum iterations. maxeval: SCAD parameter for maximum evaluations. fs.method: The method for feature selection. One out of scad, 1norm, scad+L2, DrHSVM. See package penalizedSVM for the methods.

Details

Do a ncv-fold crossvalidation (CV) with repeats repeats. Each grouping assignment from groupings will be used as classification task. If more than one grouping is present, the package uses parallelisation in the package multicore.

This function is usually called by doCV.

Value

CV result. A list with three elements res, featlist and performance, holding the crossvalidation data, an extracted features list in each cv-iteration and an overall performance object, holding information on the performance (AUC values and a roc curve object that can be plotted). See resultCV for making a summary plot for the CV.

Author(s)

Christian Bender ([e-mail])

References

Zhang, Hao Helen and Ahn, Jeongyoun and Lin, Xiaodong and Park, Cheolwoo: Gene selection using support vector machines with non-convex penalty. Bioinformatics (2006) 22 (1): 88-95

See Also

doCV Help of penalizedSVM package.

Examples

## Not run: 
#todo

## End(Not run)

bootfs documentation built on May 2, 2019, 5:50 p.m.