search

Waikato / meka

Multi-label classifiers and evaluation procedures using the Weka machine learning framework.
http://waikato.github.io/meka/
GNU General Public License v3.0
200 stars 76 forks source link

CRUpdateable && HoeffdingTree #57

Open Mali-DS opened 5 years ago

Mali-DS commented 5 years ago

Hi, I am trying to use CRUpdateable for Multi target regression, and I found the CRUpdateable 's default classifier is HoeffdingTree, but when I call method as buildClassifier, it gives me this error:

weka.core.UnsupportedAttributeTypeException: weka.classifiers.bayes.NaiveBayesUpdateable: Cannot handle numeric class!

my code is here 👍 CRUpdateable classifier = new CRUpdateable(); HoeffdingTree ht = new HoeffdingTree(); classifier.setClassifier(ht); classifier.buildClassifier(trainingInstances);

If the HoeffdingTree is not for regression why it is default classifier in CRUpdateable, and if it is for regression why it doesn't work?

Thanks Mali

jmread commented 5 years ago

Indeed, Meka does not explicitly support regression yet. Note that CR = Classifier Relevance, and 'buildClassifier' indicates it is building a classifier.

Best,

Jesse

Mali-DS commented 5 years ago

Hi Jesse, Thanks for your reply, As I know Meka recently added MultiTargetClassifier to support regression, and CRUpdateable is one of classes to go in this way:( also it has MultiLabelClassifier to support classification)

public class CRUpdateable extends CR implements IncrementalMultiTargetClassifier

Regards, Mali