One way is to make them one data set, train the learner on the first part,
and test it on the second.
Another is to use the "Model" method. This will learn (and return) a model
on the whole original dataset. To run that model on another dataset, you
may need to write a dummy learner whose "BuildClassifier" method just
reads in the model.
We wrote a StaticBN learner that does that -- takes a Bayesian network
from a file and applies it to your incoming data. As a degenerate case
I wrote one that doesn't learn anything, but returns the true class (when
known) during testing. (That's useful for checking that you're calculating
evaluation metrics properly, but not much else.)
recently i saw somebody asked about some way to apply a
from some training data, to new data for which the class atribute is
unknown. I never saw an answer. I would like to know if someone knows a
way to do this. It would be very useful for my investigation, otherwise i
would have to develop some tool to do that for me.
Charles R. Twardy www.csse.monash.edu.au/~ctwardy
Monash University sarbayes.org
Computer Sci. & Software Eng. +61(3) 9905 5823 (w) 5146 (fax)
Allow the president to invade a neighboring nation, whenever he shall
deem it necessary to repel an invasion, ... and you allow him to make
war at pleasure. --Abraham Lincoln