Want to use SVM to classify inputs into groups

I have a 4 sets of data which i wish to classify into 4 groups. each group have 100 sets of data.

This can be done easily using nprtool, where i load the input, assign the target, and it train and give me the classification accuracy. i can then give new data to test and see the classification accuracy

But i'm having trouble doing it with svm. i've read numerous concept of svm, about the hyperplane, kernel function etc. but when comes to the actual implementation, not much information or working example on it. matlab built in help file only shows example for binary classification using

SVMstruct=svmtrain(input,target,'kernel function','rbf');
newClasses=svmclassify(SVMstruct,test_input)

There are many .m file available online, but they are never as simple as: assign input, assign target, train, insert test input, receive classification output

After looking around, i found a toolbox from http://cmp.felk.cvut.cz/cmp/software/stprtool/ it's a toolbox for matlab with svm inside. it has a GUI demo, but after loading data, it will have this error when training.

Undefined function 'smo_mex' for input arguments of type 'double'.

Error in smo (line 90)
[model.Alpha, model.b, model.nsv, model.kercnt, model.trnerr,
model.margin]...

Error in demo_svm>svm_train (line 635)
     model = smo( trn, options);

Error in demo_svm (line 358)
   [Alpha,bias] = svm_train( data, trn, ker, arg, C );

Error while evaluating uicontrol Callback

From that website, it has a user manual with some example, even when i tried their examples, the same error occur, seems like the function SMO has some error and can't accept double.

Is there a way to make it work? Or is there a working .m file or toolbox that does the classification correctly without error? Please help.

Answer 1

[Ilya]

If you have 14b or later, use the fitcecoc function from the Statistics and Machine Learning Toolbox.