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 if we directly use classification learner App to train and test my 3 class problem.
Hi! May be, it is off-topic, but will be helpful for you -:). Generally speaking, there are two possibilities to perform multi-class (n) classification by means of SVM:
1. n times perform One-Vers-Rest classifications
2. n*(n-1)/2 times perform One-Vers_One classifications.
What is better for your task... You should select by means of Cross-Validation.
Dear,
Choose between various algorithms to train and validate classification models for binary or multiclass problems. After training multiple models, compare their validation errors side-by-side, and then choose the best model. To help you decide which algorithm to use, see Train Classification Models in Classification Learner App.
https://www.mathworks.com/help/stats/classification-learner-app.html
As Sergey pointed, normally you can divide into two different classifications (one vs one or one vs the rest).
In Matlab, you can use: https://se.mathworks.com/help/stats/classificationecoc-class.html
and then change the "Coding Design" (check the definitions in that page), one vs one, one vs all, but there are other possible designs.
- Badges
- Science topic