MLWizard Command Line Interface

back to main page

MLWizard can be used via the command line, as well. It will be started using the provided jar-file:

java -cp /path/to/rapidminer.jar:/path/to/MLWizard.jar de.dfki.madm.mlwizard.cli.CommandLineInterface <dataset> <task> [outfile]

The first argument is the path to the dataset in XRFF-format that will be used as input.

The second argument is the task that should be performed and can be one of the following:

  • wizard - runs the complete wizard
  • metafeatures - computes the meta-features
  • recommend - recommends classifers
  • evaluate - evaluates all classifiers
  • construct - constructs the classification system

The third argument is optional and defines where the results are written to. The result depends on the task and might be meta-features, a parameter set or a RapidMiner pipeline. If no file is supplied, the results are written to stdout.

A sample session might be:

java -cp /path/to/rapidminer.jar:/path/to/MLWizard.jar de.dfki.madm.mlwizard.cli.CommandLineInterface iris.xrff wizard result.xml
[ 0] k-NN               0.87
[ 1] Rule Induction     0.87
[ 2] Neural Net         0.87
[ 3] SVM                0.86
[ 4] Decision Tree      0.84
[ 5] Naive Bayes        0.79
Enter comma separated numbers of classifiers you want to be evaluated:
[ 0] k-NN               0.97                                                                                        
          k-NN.weighted_vote           true                                                                         
[ 1] Rule Induction     0.96                                                                                        
          Rule Induction.criterion     information_gain                                                             
[ 2] Neural Net         0.98                                                                                        
          Neural Net.decay             false                                                                        
[ 3] SVM                <not evaluated>                                                                             
[ 4] Decision Tree      <not evaluated>                                                                             
[ 5] Naive Bayes        <not evaluated>                                                                             
Enter number of classifier you want a system for:

The created RapidMiner pipeline will be stored in result.xml and can be imported into Rapidminer.

back to main page

Last modified:: 05.07.2012