I recently read this simple introduction to Adaboost as a review after learning about it a few years ago. This is in preparation for actually using it to solve a problem I am working on which is slightly different than anything addressed in this paper or that I have found online. Now in the problem I am trying to solve there is a continuous scale, rather than the binary problem addressed in the paper (and almost everywhere online) or even the distinct multistate (kind of like an enum) problems they mention at the end of the paper. Imagine experts scoring some performance on a continuous scale between 1-10 where scores like 8.4 mean twice as good as 4.2.

The way that they did the multistate version in the paper was to break it into a series of binary tasks but this completely defeats the point since it throws out a lot of the information about what the scale means and just makes it a named class. I was wondering what methods people know of for this continuous problem and if an adaboost idea can still be used. thanks.


1 Answer 1


What you are describing is called "Ordinal Regression" (see e.g. here).

There is a boosting algorithm which I think was derived from Freund's RankBoost algorithm for ranking, called ORBoost.

You might find, however, that Ordered Logistic Regression is easier to implement for your problem.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.