Exception while running a classification

[ecvgit]

I am running a classification problem. The input is a file in ARFF format. The file works fine in Weka. The class distribution is a bit skewed. Is this a bug?

Traceback (most recent call last):
  File "/usr/local/bin/run_experiment", line 9, in <module>
    load_entry_point('skll==1.1.1', 'console_scripts', 'run_experiment')()
  File "/usr/local/lib/python2.7/dist-packages/skll/utilities/run_experiment.py", line 108, in main
    ablation=ablation, resume=args.resume)
  File "/usr/local/lib/python2.7/dist-packages/skll/experiments.py", line 848, in run_configuration
    _classify_featureset(job_args)
  File "/usr/local/lib/python2.7/dist-packages/skll/experiments.py", line 446, in _classify_featureset
    grid_jobs=grid_search_jobs)
  File "/usr/local/lib/python2.7/dist-packages/skll/learner.py", line 1417, in cross_validate
    stratified else KFold(len(examples.labels),
  File "/usr/local/lib/python2.7/dist-packages/sklearn/cross_validation.py", line 441, in __init__
    test_split = test_split[test_split < len(label_test_folds)]
TypeError: object of type 'numpy.int64' has no len()

[aoifecahill]

Could you share the configuration that you used?

[ecvgit]

[General]
experiment_name = COMP_CV
task = cross_validate

[Input]
train_directory = train
featuresets = [["outall_fixed_wv_with_u_removed.arff"]]
learners = ["RandomForestClassifier", "DecisionTreeClassifier", "SVC", "MultinomialNB"]
label_col = risk_type
id_col = ID

[Tuning]
grid_search = true
objective = accuracy

[Output]
log = output
results = output
predictions = output

[aoifecahill]

Hmm, I've been able to run a cross validation experiment with some of the sample weka files and did not see that exception.

How skewed is your distribution? I'm not sure that this would even cause that exception though.

[desilinguist]

@ecvgit are you still experiencing this issue?

[desilinguist]

I am going to close this for now. Please feel free to comment if you are still experiencing this issue and I will reopen.