Can we use classification learner App in Matlab for Multi class problem?

It is obvious that we can use it for binary classification having 2 classes.What if we have 3 classes?.I have seen functions like SVMtrain,FITCSVM,SVMCLASSIFY in Matlab. But,i was wondering what if we can directly use classification learner App to train and test my 3 class problem.

[Greg Heath]

Sorry, I am not familiar with it

Greg

Answer 1

[Baptiste Ottino]

Yes, you can. I know because I played with Fisher's iris data set a little bit, and I can tell you that the App is capable of solving multi-class problems.

EDIT

Sorry, my answer was a bit short. For Fisher's iris data set:

 load fisheriris
 fisherTable = table(meas, species)

For solving multi-class problems in the App, the best way is to use a table, with all your predictors as the first p columns, and the labels as the last (in that case, the labels are strings).

Open the App, select 'New session', 'From the workspace'. Select fisherTable in Section 1, and Section 2 should look like this:

So the meas are predictors, and the species are the response. Choose your validation method in Section 3 and you should be good to go! This should work for any 3-class data set.

[Greg Heath]

How do I get one?

Greg

[Baptiste Ottino]

Edited my answer to be more precise.

[Giansu]

Hi,

I wonder to know how the multi-class classification internally works in the "classification learner" app.

Are binary classifiers in one-versus-one or one-versus-all configuration? Or are they arranged with a different approach?

I do not find this information in the documentation.

Thank you.