> Is it possible to specify a single set of training data to use as the source
> classifier for several different test data sets in the experimenter? I’m
> attempting to see how well classifiers from a particular data sources
> generalize to data sets generated from other sources. From the basic
> interface, this doesn’t seem possible. I’m not quite sure how to specify it
> if it is in the advanced.
With a bit of fiddling around, you can. The developer version now
allows you to use train/test sets (just grab a snapshot of the
developer version from the Weka homepage). There is no way of
specifying the test sets in the GUI themselves, the test set filename
is determined using the training set's relation name. Chapter "5.2.2.4
Other Result Producers" of the WekaManual, section "Explicit Test-Set
Result Producer" explains this result producer a bit (how the filename
of the test set is determined).
Now, in order to use this result producer in your environment, you
need to create copies of your training set. Use a different relation
name for each of them (e.g., by appending a number), that the correct
test set can determined in the Experimenter.
Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/ Ph. +64 (7) 858-5174
_______________________________________________
Wekalist mailing list
Send posts to:
Wekalist@...
List info and subscription status:
https://list.scms.waikato.ac.nz/mailman/listinfo/wekalistList etiquette:
http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
