Question

I gave a matrix with labels as 1,2,3 to libsvm and got 70% accuracy… I was wondering if it is correct or not ?

Does anyone know if libsvm handles multi class by itself and if it does is it one v/s all ? If not what exactly is this 70% accuracy ?

Answer 1

The official documentation for LibSVM states

LIBSVM implements the “one-against-one” approach (Knerr et al., 1990) for multi-
class classification. If k is the number of classes, then k(k − 1)/2 classifiers are
constructed and each one trains data from two classes.

The document goes on to say that classification is essentially the outcome of a voting procedure. Each of the classifiers votes on each data point and it is assigned to whichever class has the most votes for it. For accuracy, I presume these labels would be matched against the known class labels and the percentage reported is the percentage of label agreement.