|
I've a text classification problem in which there are four classes. It is a multiclass problem but not a multilabel one. I'm using sklearn.svm.LinearSVC with default value of multi_class. So by default it is acting as one vs rest classifier. In this case does it create four different one vs rest classifiers for four classes and then take some sort of average of that? or if it is using one of those four classifiers finally, is it the one that optimizes the class with label 0 (I have labels as 0, 1, 2, 3)? Also, I don't understand that if each classifier has an option to be used as one vs rest classifier then what is sklearn.multiclass.OneVsRestClassifier? This is not a model for sure, as it is written as "One-vs-the-rest (OvR) multiclass/multilabel strategy". |
