One vs one SVM for 10 classes


I want to use soft margin SVM for my dataset. My dataset contains 10 digits - 0 to 9.

I need to train using one-vs-one SVM classifier(one digit is class +1 and another digit is class -1) for digits ‘1’(+1) and ‘5’(-1).

Do I need to take subset from the training and testing dataset ?

Or should I test the SVM model on all the classes.


you can try one vs rest classification method using SVM technique.