Is there a way to perform Attribute selection(feature selection) regardless of method only for the training set before passing data for Cross Validation using the Weka API?
I currently think that the only possible way to perform this using the Weka API is through AttributeSelectedClassifier. However I am not yet sure whether this method performs attribute selection first in the whole dataset (without taking into account the cross-validation folds) and then classification, thus possibly introducing bias into the cross validation evaluation result.
Any Ideas?
[link][comment]
