what i done until was:
- evaluating two 100% equal test-setups with the same data files in weka
explorer and knowlodge flow
They're not 100% the same, see below.
- re-evaluating the model on my dataset to-classify
=> what i expectet were 2 equal outputs, what i get for real were 2
i use 10x CV in both cases.
my knowldege flow setup
is there anything i have to consider
flow by building my model with CV?
You're using the CrossValidationFoldMaker to create test sets but use an
explicit test set on all the, I assume, 10 folds of the
cross-validation. It doesn't surprise me that you get different results
(your train/test set pairs are different from the Explorer). If you use
the CVFoldMaker, you have to link both connections, trainingSet AND
testSet, to your classifier.
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
Ph. +64 (7) 858-5174