I'm trying to train multi label classifiers in scikit-multilearn through the MEKA wrapper. In MEKA, after selecting your multi label classifer, you select a WEKA single label classifier to implement it with.
In the MEKA GUI, you can also alter the settings of the single label WEKA classifier (as shown in the image below). However, I can't find a way to alter those settings through the Java terminal commands, and thus also not through the MEKA Python wrapper.
You set up a MEKA classifier through the wrapper as follows:
meka = Meka(
meka_classifier = "meka.classifiers.multilabel.BR", # Binary Relevance
weka_classifier = "weka.classifiers.trees.J48", # J48 classifier
meka_classpath = meka_classpath, #obtained via download_meka
java_command = '/usr/bin/java' # path to java executable
)
The image below shows the options in the MEKA GUI of the WEKA single label classifier:
MEKA GUI with WEKA classifier options
You would think you can add those options behind the weka_classifier line (e.g. weka_classifier = "weka.classifiers.trees.J48 -C 0.25 -M 2"), as that is how to add options to the MEKA multi label classifier, but none of this works. In the documentation, I can only find examples where they edit the MEKA classifier options, but not of the WEKA one.
Does anyone know how to edit the WEKA classifier options in MEKA through the Java terminal commands/scikit-multilearn wrapper? Thanks in advance!
--What I've tried--
I read the documentation, but it only shows examples where MEKA options are altered. For example, the following image shows how options are edited for meka.classifiers.multilabel.PS, but (as all other examples) not for the WEKA classifier:
Example of MEKA command options
I tried to edit the WEKA line of the wrapper in a lot of ways (with --, with -- \ , without all of that, etc.) None of it has worked so far.
