Approach state of art - Githubissues

jekyll / classifier-reborn

A general classifier module to allow Bayesian and other types of classifications. A fork of cardmagic/classifier.

https://jekyll.github.io/classifier-reborn/

GNU Lesser General Public License v2.1

554 stars 110 forks source link

Approach state of art #158

Open arjunmenon opened 7 years ago

arjunmenon commented 7 years ago

How do you include the TF-IDF weights in this method? Compared to simple MNB having count of plain bag of words, MNB with TF-IDF gets more accuracy. How do you implement this? www.cs.waikato.ac.nz/~eibe/pubs/kibriya_et_al_cr.pdf

Also, as a bonus a complement naive bayes further improves the problem of inconsistent dataset size, which with plain MNB favours the larger one. This scenario is fairly common. Some inputs on that as well would be appreciated,.

Ch4s3 commented 7 years ago

We've discussed TF-IDF, but I haven't had time to dig into it.