I am classifying some data as a dummy-test against a zero vector, using a Support Vector Machine (SVM), as follows:
kernel = 'linear'; C =1;
class1 = double(data(labels==1,:));
class2 = zeros([size(class1,1),size(class1,2)]);
data = [class1;class2];
theclass = [ones(size(class1,1),1); -1*ones(size(class2,1),1)];
%Train the SVM Classifier
cl = fitcsvm(data,theclass,'KernelFunction',kernel,...
'BoxConstraint',C,'ClassNames',[-1,1]);
% Cross-validation of the trained SVM
CVSVMModel = crossval(cl)
Where can I retrieve the performance of these classifications, as for instance classification accuracy, from crossval?
Edit: I am also wondering, how this kind of cross-validation works, since it is applied to an already fully trained SVM? Does it take the full dataset, and partitions it into (e.g.) 10-folds and trains new classifiers? Or is it then only predicting on the 10 test-sets?
