I'm trying to run a classifier with the InputMappedClassifier, since I know the test arff is missing some attributes in the training arff. However, when I run:
java -cp ./weka.jar weka.classifiers.misc.InputMappedClassifier -t aa/lang-train.arff \
-T aa/lang-test.arff -W weka.classifiers.trees.J48 -classifications \
weka.classifiers.evaluation.output.prediction.PlainText
It generates the exception:
java.lang.IllegalArgumentException: Invalid class index: 2466
at weka.core.Instances.setClassIndex(Instances.java:1293)
at weka.core.converters.ConverterUtils$DataSource.getStructure(ConverterUtils.java:346)
at weka.classifiers.evaluation.output.prediction.AbstractOutput.printClassifications(AbstractOutput.java:625)
at weka.classifiers.evaluation.output.prediction.AbstractOutput.print(AbstractOutput.java:702)
at weka.classifiers.evaluation.Evaluation.evaluateModel(Evaluation.java:1572)
at weka.classifiers.Evaluation.evaluateModel(Evaluation.java:649)
at weka.classifiers.AbstractClassifier.runClassifier(AbstractClassifier.java:297)
at weka.classifiers.misc.InputMappedClassifier.main(InputMappedClassifier.java:943)
If I run it without -classifications, it works. How can I get the classifications?
You're giving the InputMappedClassifier the wrong options. It's complaining that you are giving it the training (-t) and test (-T) data. It supports the following:
Options specific to weka.classifiers.misc.InputMappedClassifier:
-I
Ignore case when matching attribute names and nominal values.
-M
Suppress the output of the mapping report.
-trim
Trim white space from either end of names before matching.
-L <path to model to load>
Path to a model to load. If set, this model
will be used for prediction and any base classifier
specification will be ignored. Environment variables
may be used in the path (e.g. ${HOME}/myModel.model)
-W
Full name of base classifier.
(default: weka.classifiers.rules.ZeroR)
-output-debug-info
If set, classifier is run in debug mode and
may output additional info to the console
-do-not-check-capabilities
If set, classifier capabilities are not checked before classifier is built
(use with caution).
So your command should look like this:
java -cp ./weka.jar weka.classifiers.misc.InputMappedClassifier -W weka.classifiers.trees.J48 \
-t aa/lang-train.arff \
-T aa/lang-test.arff \
-classifications weka.classifiers.evaluation.output.prediction.PlainText