machine learning - algorithm for building multi-class (versus binary) classifiers -
i student , doing research text classification. have read several paper that. found many people using naive bayesian classifier.
i have 4 class classify. , read svm can classify thing 2 class,..yes/no 1/0
is there algorithm besides nbc algorithm building classifiers separate data more than 2 classes?
most ml techniques capable of building multi-class classifiers, instance:
decision tree (eg c4.5)
neural network via softmax (aka multi-layer perceptron, or mlp)
lda (linear discriminant analysis)
naive bayes
support vector machines routinely used multi-class classification (see example excellent scikit-learn library), using instance, "one-against-many" inductive approach. in other words, data trained on first svm separate data class versus else. "everything else" data passed second svm separates data class ii versus else, , on.
Comments
Post a Comment