0

Can AIC/BIC be used to compare logistic regression against a random forest (RF) model? I've seen discussion where AIC/BIC is more apt for linear models and not RF.

user19226726
  • 11
  • Hi again, user19226726 :-) I recommend you to edit the question a bit and include a link to the discussion. – Ute Aug 26 '23 at 15:10
  • 1
    Does this https://stats.stackexchange.com/questions/486566/akaike-information-criteria-applied-on-random-forest thread answer your question?

    @Tim 's answer agrees with my intuition that neither the log likelihood nor the number of parameters is well defined for a random forest.

     – Peter Flom Aug 26 '23 at 15:28
  • I think it does - but there are some journal articles that DID use AIC/BIC for RF. So that made me question it. Are there any scholarly resources that confirm AIC/BIC isn't applicable for RF? I couldn't find any. – user19226726 Sep 09 '23 at 16:14

0 Answers0