Posted by Bobby Dazzler on December 06, 19101 at 23:12:07:
I am trying to test boosting algorithms on my dataset which i am currently classifying with C4.5 to see the improvement in performance.
The dataset i am using is multiclass and multidimensional (all parameters are continuous), is AdaBoost alright to use? Should i use AdaBoost MH or M2 instead?
I see that the regularised version AdaBoost-reg is only for use in MatLab which i unfortunately cannot use. I also see that Schapire's new BoosTexter is something that I would very much like to try but i really need to test what effect the extension of boosting on my C4.5 tree.
Is there a version of AdaBoost that is easily accessible to build on top of another classifier? I have read many papers on the use of this but i have no idea how to do this! Where could i get this algorithm from? Does it exist as an executable or does it have to be patched into C4.5?
I will be very grateful of your help