rfe function from caret to perform svm-rfe feature selection
1
0
Entering edit mode
13 months ago

Hi everyone, I'm trying to perform an SVM-RFE feature selection using the caret package on R: I have a dataset with 1000 and more features (miRNA expression counts, normalized) as columns (+ one column with the class, normal vs tumor) and few hundreds of samples as rows. I've found that the rfe function can be used for this purpose, but I'm not sure how to set the arguments. In particular I cannot understand what does the sizes argument mean: just some features will be used for the model? does it make sense when I need to use RFE for selecting them? And what does happen if I do not set it? I didn't understand this point.

I tried to set the rfeControl first, then, I set the rfe, but I don't know if this is the correct way to do it

    control <- rfeControl(functions = caretFuncs, #It is correct for svm-rfe?
method = "repeatedcv",
number = 10,
verbose = FALSE)

recursivefs <- rfe(dataset[,-1], dataset\$class,
method="svmLinear",
rfeControl = control)


I set the argument method="svmLinear" as I found that it determs the model used, but I'm not sure about this; moreover, I don't know what to set on "sizes". Can anyone help? I'm at the very beginning with feature selection, so every advice will be very appreciated. Thank you.

R feature selection caret rfe svm • 1.2k views
0
Entering edit mode
9 months ago
tzehuey85 • 0

Hi @lenC_biotecLover pls refer to webiste

http://www.feat.engineering/recursive-feature-elimination.html