Distributed RapidMiner

[Michael Wurst]

Hi all,

I am co-developer of RapidMiner (formerly Yale, 
http://www.rapidminer.com). Isa Drost pointet the Mahout project out to 
me. We are currently working on a distributed version of RapidMiner that 
will run in a cluster. The focus is, however, very different from the 
one of Mahout. We are focusing on distributed databases and highly 
complex tasks, such as evolutionary computing for feature selection or 
parameter optimization.  These task require to perform many training and 
evaluation cycles on variants of the same data mining task (e.g. using 
different feature sets). Therefore, we decided not to use map/reduce for 
this kind of application but some more traditional methods of 
distributed data mining, task distribution and scheduling.

Creating a map/reduce based data mining library/system is sure highly 
relevant. I'm looking forward to the first results of Mahout! If you 
need any assistance, let me know.

Cheers,
Michael