Thank you for your answer, but I am afraid that caret uses kernlab for executing LS-SVM, and only for classification (in fact, caret is a package that does not create models by itself but using other packages), as you can see in
Regression can be performed using svm function from e1071 package. It is an interface to libsvm. You can tune kernels, regularization parameters, run classification or regression.
Here is an overview (from 2006) of SVM packages in R:
Karatzoglou A. et al, Support Vector Machines in R, Journal of Statistical Software, April 2006, Volume 15, Issue 9., http://www.jstatsoft.org/v15/i09/paper
Thank you very much for your answer. Yes, e1071 package is a good choice for SVM regression (in fact I use that package for my works in SVM regression), but I am looking for an implementation of LS-SVM in R, and not the (say) "classical SVM" which e1071 works with. For the differences between them, you can see, for example,
Anyway, there are lots of papers and tutorials on LS-SVM in the internet, if you need more information.
There are some implementations of LS-SVM for MATLAB but, as I work with the R environment, I was looking for an specific package for R, which I can not find (may be it does not exist).
You are looking for kernlab::lssvm ? You can use lssvm() from kernlab package. Besides, you can apply a kernel out of 8 kernel functions such as rbfdot, polydot, vanilladot etc.
Please anyone can help me on the R implementation of this new algorithm SCUT as an extension of SMOTE method for umbalanced distribution with multiclass response variable and thank you
As mentioned, kernlab does not perform regression via LS-SVM. Another package called liquidSVM exists that performs LS-SVM, though it needs polishing regarding outputting the model coefficients (details about indexes of data used in each of the n-fold training steps, final selected value of model parameters such as gamma and cost after grid search and optimization, model intercept value). Although this is not a problem if you wish to use the package itself to calculate the output values for your data and you don't need to export the model to another software.
You can fit lssvm with several kernels (linear, polynomial or radial basis function) using the kernlab package. If you like to use caret, you can access these via the lssvmLinear, lssvmPoly and lssvmRadial methods.