MATLAB: I am going to classify multispectral remote sensing image using SVM .

Question

I have written a code for classifying three crops using SVM training , The three crops are cotton ,wheat and gram. Now to get more accuracy I want to optimize the training data using genetic algorithm then I wish to train the optimized data using SVM train and then want the classification result . First want to compare the result with simple SVM algorithm , I am attaching the codes here please help me to correct that code so that i could get the better kappa coefficient .ore if u suggest some other parameters of comparison of two techniques . One more thing is please suggest me how to use kernels in SVMtrain instruction ..!

Answer 1

The following section in the documentation shows how you can use cross validation to tune your svm parameters using optimization:

http://www.mathworks.com/help/stats/support-vector-machines-svm.html#bsr5o1q

Go through the examples and how-to's in this link, that should answer most of your questions:

http://www.mathworks.com/help/stats/support-vector-machines.html