[R] R Weka and cobweb

Hans W. Borchers hwborchers at yahoo.de
Sun Aug 12 13:55:25 CEST 2007


 <adschai <at> optonline.net> writes:

> 
> Hi,
> 
> I never use cobweb before and I'm quite new to this. I have a couple of
questions around the cobweb
> implementation in R Weka. If you could supply answer or insight, I would
really appreciate.
> 
> 1. From Fisher's paper in 1987, it seems that Cobweb only deals with nominal
data. In R Weka cobweb, is it
> allowed to accommodate real/continuous value?
> 2. My understanding is that Cobweb clusters based on Category Utility
function. However, if I would like to
> specify my own objective function, how can I do so?
> 3. I have about 130,000 data points (each with about 50 attributes) for
training and another same amount for
> testing. Would this cause any memory or performance problem?  
> 
> Thank you,
> 
> - adschai
> 

RWeka is simply an interface to the WEKA data mining software.
Why don't you just put your question to the Weka mailing list
at https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist?

>From my own former experience I would say that Cobweb is never
applicable to large data sets (and was not meant for that) and
for numerical attributes the results are not very meaningful.

Hans Werner Borchers



More information about the R-help mailing list