The current implementation of Random Forests looks good indeed. My latest tests on NSL-KDD (http://nsl.cs.unb.ca/NSL-KDD/) shows similar recognition rates such as those reported in the paper. After the release of Mahout 0.3 (and the end of the current code freeze), I should post some additions, for example saving the training decision forest and use it to classify test datasets. And will also post a detailed description of the results on NSL-KDD.
On Mon, Mar 1, 2010 at 7:19 PM, Ted Dunning <ted.dunn...@gmail.com> wrote: > Not at this time. We would love to have contributions. > > On the other hand, we do have an implementation of Random Forests that looks > pretty robust on initial testing. That may satisfy your needs. > > On Mon, Mar 1, 2010 at 5:45 AM, Meher Anand (JIRA) <j...@apache.org> wrote: > >> Is there any plan to implement ID3 and C4.5 algorithms in Mahout? I am an >> undergrad student who's working on Mahout as part of my senior year project. >> I plan to implement decision tree learning as part of my project and was >> wondering if it is a good enough idea to incorporate it into the mainstream >> application. >> > > > > -- > Ted Dunning, CTO > DeepDyve >
