How to use logistic regression for multi-class classification in rapidminer?

mahsa_d1992
mahsa_d1992 New Altair Community Member
edited November 2024 in Community Q&A
Hi. I have a dataset which has nominal values. Also, the column I want to use as prediction is divided into 3 classes. I do not know how to use logistic regression in this state?

Answers

  • Telcontar120
    Telcontar120 New Altair Community Member
    Logistic regression only works with binominal (2-class) labels.  However, the Generalized Linear Model operator has a multinomial regression solver which is suitable for 3-class labels (or more). 
    There is a sample process for it available in the operator help that should guide you.
    There are also many other ML operators that can handle multinomial labels.
  • varunm1
    varunm1 New Altair Community Member
    Hello @Telcontar120

    Is the solution suggested in your post similar to the multinomial logistic regression in python? I generally use python for this. I feel they might be similar but just want your insight.

    Also, we can try polynomial to binomial (One Vs All) and apply logistic regression on that directly.
  • Telcontar120
    Telcontar120 New Altair Community Member
    @varunm1 this is more of a question for @mschmitz but as far as I know, the multinomial GLM in RapidMiner would be very similar to the multinomial LR python function.
    The one-vs-all-other method is a workaround but isn't suitable if you really do want simultaneous predictions for multiple classes.

  • varunm1
    varunm1 New Altair Community Member
    edited August 2019
    Thanks Brian.