0

Is there a possibility to get the classification score for each ova estimator in a multi class SVM in scikit? The overall score of the muldi class SVM is not what i need.

To be more specific. If i have a trained multi class SVM and i use it on my test data i want not only the overall score from the multi class SVM for each sample. I also need let me say the 10 best ova classifier reults from inside the multi class SVM. The confusion matrix would maybe one possibility but i'm not sure if the matrix reflects each classification process.

theSOMguy
  • 1
  • 4
  • Please add some sample code. How are you initializing the svm? Are you setting the decision_function_shape to 'ovr' Or are you using OneVsRest with SVC? – Vivek Kumar Jan 31 '18 at 13:05
  • If you are using only SVC, then even though it does what you describe but we cant access those in python, because scikit delegates all the functionality to underlying Libsvm classes. But if you are using OneVsRestClassifier, then you can access the `estimators_` attribute to access all the binary classifiers. Post your code and we can have a working example based on that. – Vivek Kumar Jan 31 '18 at 13:16

0 Answers0