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

Prediction results #8

Closed niedakh closed 8 years ago

niedakh commented 8 years ago

Hi,

I'm trying to get prediction results after testing from meka. I'm loading a prebuilt classifier: /usr/bin/java -cp "/home/niedakh/scikit/meka/meka-release-1.9.0/lib/*" meka.classifiers.multilabel.LC -W weka.classifiers.bayes.NaiveBayes -threshold 0 -verbosity 5 -t ~/engine/scikit-multilearn/meka/data/scene-test.arff -T ~/engine/scikit-multilearn/meka/data/scene-test.arff -l classifier.dump

With verbosity 5 I am getting results such as: | 1196 [5] [0, 1, 2, 3, 4, 5]

what I get from meka.core.Result is that the first is the truth from the test set, the right are the predictions?

With verbosity 6 I am getting:

| 1196 [ 0 0 0 0 0 1 ] [ 0,0 0,0 0,0 0,0 1,0 0,0 ]

Which would suggest that with verbosity 5 meka predicts all labels, with verbosity 6 it predicts only the fifth one. Is this a bug in verbosity 5?

niedakh commented 8 years ago

nvm, managed to make it work.