As others have commented, there are many libraries implementing multi-class SVMs natively (personally, I recommend LibSVM, as in Olarik Surinta's answer).

However, if you do not want to use additional external software, you can easily construct a multi-class SVM starting from the binary SVM (actually, this is what most of the external libraries do). You can do this using different methods, the most popular being one-versus-all classification [1]. The idea is simply to train M binary SVMs, where M is the number of classes, and in the testing phase to output the class corresponding to the largest output between the M classifiers. You can implement this easily in Matlab. I think there is also some ready-to-use code on Matlab Central.

Once you become familiar with one-versus-all classification, there are other approaches you can additionally try, see [2].

[1] Rifkin, Ryan, and Aldebaro Klautau. "In defense of one-vs-all classification." The Journal of Machine Learning Research 5 (2004): 101-141.

[2] Hsu, Chih-Wei, and Chih-Jen Lin. "A comparison of methods for multiclass support vector machines." Neural Networks, IEEE Transactions on 13.2 (2002): 415-425.

http://www.loria.fr/~lauer/MSVMpack/MSVMpack.html

www.cmpe.boun.edu.tr/~ethem/files/papers/iwann99.pdf

www.researchgate.net%2Fpublication%2F2854340_LS-SVMlab_a_MATLABC_toolbox_for_Least_Squares_Support_Vector_Machines

Normally, I used LIBSVM for multi-class classification.

It supports MATLAB as well.

http://www.csie.ntu.edu.tw/~cjlin/libsvm/

See also "Cascading SVMs as a Tool for Medical Diagnosis using Multi-class Gene Expression Data", International Journal of Artificial Intelligence Tools, vol. 15, no. 3, pp. 335-352, http://www.worldscinet.com/ijait/15/1503/S0218213006002709.html, World Scientific, 2006.

As others have commented, there are many libraries implementing multi-class SVMs natively (personally, I recommend LibSVM, as in Olarik Surinta's answer).

However, if you do not want to use additional external software, you can easily construct a multi-class SVM starting from the binary SVM (actually, this is what most of the external libraries do). You can do this using different methods, the most popular being one-versus-all classification [1]. The idea is simply to train M binary SVMs, where M is the number of classes, and in the testing phase to output the class corresponding to the largest output between the M classifiers. You can implement this easily in Matlab. I think there is also some ready-to-use code on Matlab Central.

Once you become familiar with one-versus-all classification, there are other approaches you can additionally try, see [2].

[1] Rifkin, Ryan, and Aldebaro Klautau. "In defense of one-vs-all classification." The Journal of Machine Learning Research 5 (2004): 101-141.

[2] Hsu, Chih-Wei, and Chih-Jen Lin. "A comparison of methods for multiclass support vector machines." Neural Networks, IEEE Transactions on 13.2 (2002): 415-425.

Thanks all I am greatly enlightened that there is possibility any many approaches and research papers, code available

Thanks Mr. Simone Scardapane

PRTools could be an option as well:

http://prtools.org

I suggest you to search on Mathworks.com/MatlabCenteral

you can find many standard useful Matlab codes for your project, some might be added to your software and some not, but available for download.

Can anyon send me the link for multiclass SVM matlab code?

Please explain the procedure for multiclass classification using LIBSVM libraray.

this link contains another tool for multiclass SVM

http://www.loria.fr/~lauer/MSVMpack/MSVMpack.html

Can LIBSVM use for multi-dimentional problems? Can any one guid me how to do that in Matlab.

Dear Isuru,

You may try fitrsvm function of matlab as well.

https://www.mathworks.com/help/stats/fitcsvm.html

matlap function

http://www.mathworks.com/matlabcentral/fileexchange/39352-multi-class-svm?focused=3776247&tab=function