I wrote a custom classifier for WEKA, and it needs to provide information besides just a predicted class for an instance. Specifically, it is a list describing how great the effect of each of a subset of the training instances was on the predicted class.
The additionalMeasureProducer interface will not work, since I need to print a string to the GUI. I would like the user to be able to choose to output this information. I had hoped that the option "Output model" in the "Classifier evaluation options" dialog box would do the trick, but WEKA ends up calling my classifier's toString() method before classification has happened. I consider this a bug, since my classifier is in the weka.classifiers.lazy package, meaning that it isn't built until the Instance to be classified is given to it.
Does anyone have any idea how to add more GUI output capabilities to a custom WEKA classifier?