R help - Jun 2010 - Cforest and Random Forest memory use

Hi all,

I'm having great trouble working with the Cforest (from the party package)
and Random forest functions.  Large data set seem to create very large model
objects which means I cannot work with the number of observations I need to,
despite running on a large 8GB 64-bit box.  I would like the object to only
hold the trees themselves as I intend to export them out of R.  Is there
anyway, either through options or editing out code and recompiling them, I
can reduce their footprint?

I've had a look at the cforest code and the culprit is the
'emsemble' area
of the object.  I suspect this part of the object contains something related
to the number of observations (I have savesplitstats set to FALSE so this
shouldn't be the issue).

Thanks,
Matt

	[[alternative HTML version deleted]]

You say you're on a 64-bit box, but are you running 64-bit R?

   -Peter Ehlers

On 2010-06-10 4:36, Matthew OKane wrote:
> Hi all,
>
> I'm having great trouble working with the Cforest (from the party
package)
> and Random forest functions.  Large data set seem to create very large
model
> objects which means I cannot work with the number of observations I need
to,
> despite running on a large 8GB 64-bit box.  I would like the object to only
> hold the trees themselves as I intend to export them out of R.  Is there
> anyway, either through options or editing out code and recompiling them, I
> can reduce their footprint?
>
> I've had a look at the cforest code and the culprit is the
'emsemble' area
> of the object.  I suspect this part of the object contains something
related
> to the number of observations (I have savesplitstats set to FALSE so this
> shouldn't be the issue).
>
> Thanks,
> Matt

Also, you have not said:

 - your OS
 - your version of R
 - your version of party
 - your code
 - what "Large data set" means
 - what "very large model objects" means

So... how is anyone suppose to help you?

Max