"Bagging regression trees: Bagging and W-M5P"

nicugeorgian

Hello,

How can I apply bagging to the M5P algorithm, i.e., how can I use the operators Bagging and W-M5P together?

I use W-M5P to predict, by means of a regression tree, a numerical label based on polynomial and numerical attributes.

If I use W-M5P as an inner operator of the Bagging operator,

<operator name="Bagging" class="Bagging" expanded="yes">
         <operator name="W-M5P" class="W-M5P">
                <parameter key="M"	value="10.0"/>
                <parameter key="N"	value="true"/>
                <parameter key="R"	value="true"/>
         </operator>
</operator>

then the process complains, saying that Bagging is not able to support numerical labels 

Any ideas?

Thanks,
Geo

IngoRM

Hi Geo,

the reason why this does not work is quite simple: in the current version, bagging does not support numerical labels (regression problems) but classification tasks only. I have just changed that and numerical labels are now also supported by bagging. You can access this imrpoved version via CVS (described here: http://rapid-i.com/content/view/25/48/ ). Users of the RapidMiner Enterprise Edition will get this improvement with the next automatic update. And of course this change will also be part of the next release.

Cheers,
Ingo