Re: Important features for each fold

This is also possible in the Explorer now, by ticking "Output models for
training splits" under "More options..." in the "Classify" tab. Running the
AttributeSelectedClassifier in this way, with ZeroR as the base classifier,
would the fastest way to see all the attribute subsets for each fold.

Cheers,
Eibe

On Tue, 14 Feb 2023 at 16:28, Michael Hall <[email protected]> wrote:

>
>
> > On Feb 13, 2023, at 9:09 PM, Peter Reutemann <[email protected]>
> wrote:
> >
> >> How can we get features selection for each fold of the cross
> validation. I think it was possible in the earlier version of weka explorer
> but I cannot see the option now.
> >
> > I'm not aware of such an option ever being available. The Attribute
> > selection tab in the Weka Explorer only gives you the summary of how
> > often an attribute was selected during cross-validation (when using
> > subset evals) or the average rank/merit when performing ranking.
> >
> > For your desired output, you would have to simulate cross-validation
> > yourself, generate train/test splits, perform attribute selection on
> > the train split and record the selected attribute indices.
> > That's something you can do with Jython, Groovy or python-weka-wrapper3.
> > ADAMS comes with a flow that performs repeated CV on a user-selected
> > dataset, populating a table with information on whether an attribute
> > was selected or not
> > (adams-weka-attribute_selection_simulated_cv.flow).
> >
>
> I thought I remembered something being available per fold with
> KnowledgeFlow but could be mistaken.
>
> _______________________________________________
> Wekalist mailing list -- [email protected]
> Send posts to [email protected]
> To unsubscribe send an email to [email protected]
> To subscribe, unsubscribe, etc., visit
> https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
> List etiquette:
> http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
>

_______________________________________________
Wekalist mailing list -- [email protected]
Send posts to [email protected]
To unsubscribe send an email to [email protected]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html