imbalanced-learn 0.7.0 is out

imbalanced-learn, probably, is your favorite python package that offers a number of re-sampling techniques commonly used in datasets showing strong between-class imbalance.

This is release should be be fully compatible with the latest version of scikit-learn.

Maintenance

• Pipelines can cope with older versions of joblib.
• Common tests have been refactored.
• Feature warnings have been removed.
• Imposing keywords only arguments as in scikit-learn.

Changed models

• imblearn.ensemble.BalancedRandomForestClassifier is expected to give different results for the same input (using the same random state).
• Fix make_index_balanced_accuracy which was unusable due to the latest version of scikit-learn.
• Raise a proper error message when only numerical or categorical features are given in imblearn.over_sampling.SMOTENC.
• Fix a bug when the median of the standard deviation is null in imblearn.over_sampling.SMOTENC.

Bug fixes

• min_samples_leaf default value has been changed to be consistent with scikit-learn ### Enhancements
• The classifier implemented in imbalanced-learn, imblearn.ensemble.BalancedBaggingClassifier, imblearn.ensemble.BalancedRandomForestClassifier, imblearn.ensemble.EasyEnsembleClassifier, and imblearn.ensemble.RUSBoostClassifier, accept sampling_strategy with the same key than in y without the need of encoding y in advance.
• Import keras module lazily.

Installation

You can install it either by using pip

pip install imbalanced-learn -U

or by using the conda package manager

conda update imbalanced-learn

The changelog can be found here, while installation instructions, API documentation, examples and a user guide can be found here.

Happy hacking,
Chris