Orange Forum • View topic - random forest ? naive bayes?

random forest ? naive bayes?

A place to ask questions about methods in Orange and how they are used and other general support.

random forest ? naive bayes?

Postby tienza » Fri Aug 16, 2013 11:30

I have a database, my result varible is dichotomous, (no/yes). I made a logistic regresion with data, and I choose four variables with significance differences on multivariable analysis. (two are continuous and two dichotomous).
I made a nomogram (from logistic regresion), a cross-validation (test learners from naive bayes and random forest) and finally a calibration plot and ROC analysis.

My question is about what to choose for the AUC and classification accuracy, between naive bayes and random forest. All variables have not normal distribution.

Thank you so much for the software and all the support.

Re: random forest ? naive bayes?

Postby bricklemacho » Wed Oct 02, 2013 10:03

Not sure if I understand the question. If you are asking which measure is better to compare learning algorithms, then based on these papers, AUC is the better measure.

Ling, Charles X., Jin Huang, and Harry Zhang. "AUC: a better measure than accuracy in comparing learning algorithms." Advances in Artificial Intelligence. Springer Berlin Heidelberg, 2003. 329-341.

Jin Huang; Ling, C.X., "Using AUC and accuracy in evaluating learning algorithms," Knowledge and Data Engineering, IEEE Transactions on , vol.17, no.3, pp.299,310, March 2005

Hope this helps.


Disclaimer: I am not part of the Orange development team.

Return to Questions & Support