I have a training set and a testing set which I want to classify with Weka. and because my attribute is a text .. I understand that I have to apply first the stringtoword filter first. As has been described here, I have to do the following: 1- load the train and test sets and apply the filter. 2- Train the classifier using the training set and save the model. 3- Load the model to apply it on the testing set.
I my experiment I have to use the trained classifier to test multiple test sets .. so, do I have to repeat 1,2 and 3 or I can repeat the third step only ?
and do I have to replace the class in testing set with a question mark (?) or I can leave the class their and the classifier will try to predict the class anyway?
You can simply repeat the third step assuming all the testing sets have been filtered and saved with the filtered version. If you select the 'More options...' button and check the output predictions box then you do not have to replace the class with a question mark.