how to label features to train by SVM?

i have to train SURF features which are stored in a excel file..what is the best way to label? i have 96x20 (48x20 for class A and 48x20 for Class B) dimension of features which i currently stored in a excel sheet..then i made another excel sheet and assign labels as A to 48 rows and next assigned 48 B labels..then i loaded both the files through xlsread in my program..

features_train = xlsread('A_B_features');

[num,labels_train]=xlsread('A_B_labels');

SVMStruct = svmtrain(features_train,labels_train,'Showplot',true);

is this the correct way for svmtrain ?? or should i keep my features and labels into a single file? if yes..how??

Answer 1

[Akshat]

Hi @Farman Shah,

Your implementation looks correct, except just one thing; "svmtrain" has been deprecated.

As a replacement, "fitcsvm" function has been introduced. Here is an example:

features_train = xlsread('A_B_features');
[num, txt] = xlsread('A_B_labels');
labels_train = categorical(txt);
SVMModel = fitcsvm(features_train, labels_train, 'KernelFunction', 'linear', 'Standardize', true, 'ClassNames', {'A', 'B'});

Here is the documentation link for "fitcsvm":

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

Hope this helps.