Logistic Regression cannot handle polynominal label.

ArsalanTahir

Hello All,
I have used logistic regression to resolve citation classification problem but error on my label class (High, Medium, and Low), tried the "nominal to numerical" operator but still face this issue. please, help it's urgent.

BalazsBarany

Hi!

Nominal to numerical converts your attributes (the ones that build up the model) to numericals. That's fine.

However, the concept of logistic regression doesn't allow for predicting multiple classes. What you can do is predicting one class vs. all the others repeatedly and taking the prediction with the highest confidence. 

There's a meta learner for that, it's called Polynominal by Binominal Classification. You can use it as a wrapper around the actual learner (logistic regression). The operator's help contains a tutorial process so you can see how it's done.

Regards,
Balázs

BalazsBarany

Hi!

Nominal to numerical converts your attributes (the ones that build up the model) to numericals. That's fine.

However, the concept of logistic regression doesn't allow for predicting multiple classes. What you can do is predicting one class vs. all the others repeatedly and taking the prediction with the highest confidence. 

There's a meta learner for that, it's called Polynominal by Binominal Classification. You can use it as a wrapper around the actual learner (logistic regression). The operator's help contains a tutorial process so you can see how it's done.

Regards,
Balázs