Hey,
I've been working on adding ARFF support for RF classifiers. I've
posted a couple of patches and wanted to explain what I had in mind.
The larger goal here is to run any integration to generate a
dictionary/meta-data and sequence file that can be consumed directly by
the RF classifiers.
The patches I've done so far are to make RF consume meta-data in a
easier to understand format (MAHOUT-1163) and to have the ARFF
integration generate meta-data in the same format (MAHOUT-1164). To
round out ARFF integration in RF all we need is to consume sequence
files of vectors. I'm working on that...
Let me know if this sounds reasonable.
Marty