Description
Hello,
I tried to use multi-label Hoeffding Tree to build my baseline, however I gots some problems.
At first
import moa.classifiers.multilabel.MultilabelHoeffdingTree;
MultiTargetArffFileStream stream = new MultiTargetArffFileStream("mediamill.arff", "-101");
MultilabelHoeffdingTree learner = new MultilabelHoeffdingTree();
I got a warning :
[WARNING] Only 1 labels found! (Expecting 101)
(Ignoring this prediction)
It means that the defaults classifier is not a multi-label classifier ?
So I use PSUpdate classifier as the base classifier in the leaf of hoeffding tree (as in mentioned in the article) :
_import moa.classifiers.multilabel.MultilabelHoeffdingTree;
import meka.classifiers.multilabel.incremental.PSUpdateable;
MultilabelHoeffdingTree learner = new MultilabelHoeffdingTree();
MEKAClassifier PS = new MEKAClassifier();
PSUpdateable base = new PSUpdateable();
PS.baseLearnerOption.setCurrentObject(base);
PS.setModelContext(header);
PS.prepareForUse();
PS.resetLearningImpl();
learner.learnerOption.setCurrentObject(PS);
Then I got [ERROR] Failed to update classifier
weka.core.UnsupportedAttributeTypeException: weka.classifiers.bayes.NaiveBayesUpdateable: Cannot handle unary class!
The dataset "mediamill.arff" I used has the numeric attributes, and the labelset is at the end of each example.
When I load data, I add a minus sign, so there should be a problem with the connection between moa and weka when loading data?
Thank you for your help!