I am trying to classify an ARFF file where the first 15000 instances have all attributes set, and the final 1000 instances are missing the last attribute (attribute is set to unknown "?").
I want to classify those missing data and export them so that the missing attributes are set (in the same format as the input).
I tried to use Weka Explorer but I couldn't figure out how to manage what I am looking for.
Can anybody help?
If you can separate the unlabelled "?" instances into their own .arff file you can then choose the option of "Supplied test set" in Weka and pass this new file in as the test set. Then click on "More options..." and check "output predictions" this will show the label of the previously unlabelled instances. Train your classifier. Then if you visualise your data and right click on the graph you'll have the option to save this new labelled data as an .arff file!
