Meeseeks.SVM: Easy Cross-Validated SVM Binary Classification
In Beirnaert/MetaboMeeseeks: Helper functions for metabolomics analysis

Description Usage Arguments Value Author(s)

****** STILL CONTAINS BUGS (inverse ROC and PR) ****** This function quickly performs a cross-validated SVM classification on a data matrix.

Meeseeks.SVM(FeatureMatrix, GroupLabels, SampleLabels = NULL,
  nFolds = 10, nSims = 20, plot.out = TRUE, plot.type = "ROC",
  nCPU = -1, plotcol = NULL, svm.kernel = "linear",
  plottitle.extra = NULL, prediction_prob = TRUE, Nplotpoints = 501,
  ...)

`FeatureMatrix`	The matrix of Features (obtained by using the xcms::groupval function). Matrix has to have columns for features and rows for samples.
`GroupLabels`	The group labels. If not a factor a conversion will be applied.
`SampleLabels`	(optional) unique sample identifier.
`nFolds`	Number of cross validation folds.
`nSims`	Number of simulations (every simulation has different folds)
`plot.out`	Whether to print the ROC curve (default is TRUE).
`plot.type`	Type of plot output. "ROC" for receiver operator characteristic (default) or "PR" for precision-recall.
`nCPU`	The number of cores to use (default is the maximum amount available minus 2)
`plotcol`	(optional) colour to use for the plot
`svm.kernel`	The kernel to be used for the svm (default is linear)
`plottitle.extra`	Optional extra character string to be added to every plot title.
`prediction_prob`	Whether to take the values for the TRUE prediction (default) or the FALSE prediction. Choosing FALSE inverts the ROC curve.
`Nplotpoints`	The amount of points used to construct the plot.
`...`	Extra paremeters to be passed along to `svm`

A ROC plot (if plot.out = TRUE) and a list with 2 elements: 1) a data frame with the ROC plot data and 2) a matrix with the variable importance for each cross validated simulation (nFolds * nSims times).

Charlie Beirnaert, charlie.beirnaert@uantwerpen.be

Beirnaert/MetaboMeeseeks documentation built on May 20, 2019, 11:09 a.m.