Help with adaBoost

Herly

Hi everyone,

I've a confusion with adaboost, I thought that adaboost was a multi classifier, that uses different models and then it combines them to have a final answer, but I can't put more than one model in it( I was expecting something like the method Vote), so can anyone explain me how adaBoost really works?

And I've another question, I'm predicting a DB with 6 different classes, but one of them has a recall of 15%, and read that I've to create another model, just to classify that class, anyone knows how to do that?

Thanks 

B_Miner

Regarding Ada Boost, I suggest you Google it, there are a lot of references. The algorithm (I think there are multiple versions actually) does not use multiple learners in the sense you were expecting it -- instead the same learning algorithm (in the sense of the same algorithm with the same parameter settings)  is used, but repeatedly fit to a weighted dataset where the weights reflect how well the example instance was predicted by the previous fitting. Hope that makes sense. See http://en.wikipedia.org/wiki/AdaBoost and also Witten and Frank give a great account of it if you have their book (wrote WEKA). Hope that helps