Imbalanced-learn

=============

Changed models
---

The following models or function might give different results even if the
same data ``X`` and ``y`` are the same.

* :class:`imblearn.ensemble.RUSBoostClassifier` default estimator changed from
:class:`sklearn.tree.DecisionTreeClassifier` with full depth to a decision
stump (i.e., tree with ``max_depth=1``).

Documentation
---

- Correct the definition of the ratio when using a ``float`` in sampling
strategy for the over-sampling and under-sampling.
:issue:`525` by :user:`Ariel Rossanigo <arielrossanigo>`.

- Add :class:`imblearn.over_sampling.BorderlineSMOTE` and
:class:`imblearn.over_sampling.SVMSMOTE` in the API documenation.
:issue:`530` by :user:`Guillaume Lemaitre <glemaitre>`.

Enhancement
---
- Add Parallelisation for SMOTEENN and SMOTETomek.
:pr:`547` by :user:`Michael Hsieh <Microsheep>`.

- Add :class:`imblearn.utils._show_versions`. Updated the contribution guide
and issue template showing how to print system and dependency information
from the command line. :pr:`557` by :user:`Alexander L. Hayes <batflyer>`.

- Add :class:`imblearn.over_sampling.KMeansSMOTE` which is an over-sampler
clustering points before to apply SMOTE.
:pr:`435` by :user:`Stephan Heijl <StephanHeijl>`.

Maintenance
---

- Make it possible to ``import imblearn`` and access submodule.
:pr:`500` by :user:`Guillaume Lemaitre <glemaitre>`.

- Remove support for Python 2, remove deprecation warning from
scikit-learn 0.21.
:pr:`576` by :user:`Guillaume Lemaitre <glemaitre>`.

Bug
---

- Fix wrong usage of :class:`keras.layers.BatchNormalization` in
``porto_seguro_keras_under_sampling.py`` example. The batch normalization
was moved before the activation function and the bias was removed from the
dense layer.
:pr:`531` by :user:`Guillaume Lemaitre <glemaitre>`.

- Fix bug which converting to COO format sparse when stacking the matrices in
:class:`imblearn.over_sampling.SMOTENC`. This bug was only old scipy version.
:pr:`539` by :user:`Guillaume Lemaitre <glemaitre>`.

- Fix bug in :class:`imblearn.pipeline.Pipeline` where None could be the final
estimator.
:pr:`554` by :user:`Oliver Rausch <orausch>`.

- Fix bug in :class:`imblearn.over_sampling.SVMSMOTE` and
:class:`imblearn.over_sampling.BorderlineSMOTE` where the default parameter
of ``n_neighbors`` was not set properly.
:pr:`578` by :user:`Guillaume Lemaitre <glemaitre>`.

- Fix bug by changing the default depth in
:class:`imblearn.ensemble.RUSBoostClassifier` to get a decision stump as a
weak learner as in the original paper.
:pr:`545` by :user:`Christos Aridas <chkoar>`.

- Allow to import ``keras`` directly from ``tensorflow`` in the
:mod:`imblearn.keras`.
:pr:`531` by :user:`Guillaume Lemaitre <glemaitre>`.

Mainly bugfix in SMOTE NC

Bug fixes

* Fix a bug in imblearn.over_sampling.SMOTENC in which the the median of the standard deviation instead of half of the median of the standard deviation. By Guillaume Lemaitre in 491.
* Raise an error when passing target which is not supported, i.e. regression target or multilabel targets. Imbalanced-learn does not support this case. By Guillaume Lemaitre in 490.

===========

**October, 2018**

.. warning::

Version 0.4 is the last version of imbalanced-learn to support Python 2.7
and Python 3.4. Imbalanced-learn 0.5 will require Python 3.5 or higher.

Highlights
----------

This release brings its set of new feature as well as some API changes to
strengthen the foundation of imbalanced-learn.

As new feature, 2 new modules `imblearn.keras` and
`imblearn.tensorflow` have been added in which imbalanced-learn samplers
can be used to generate balanced mini-batches.

The module `imblearn.ensemble` has been consolidated with new classifier:
`imblearn.ensemble.BalancedRandomForestClassifier`,
`imblearn.ensemble.EasyEnsembleClassifier`,
`imblearn.ensemble.RUSBoostClassifier`.

Support for string has been added in
`imblearn.over_sampling.RandomOverSampler` and
`imblearn.under_sampling.RandomUnderSampler`. In addition, a new class
`imblearn.over_sampling.SMOTENC` allows to generate sample with data
sets containing both continuous and categorical features.

The `imblearn.over_sampling.SMOTE` has been simplified and break down
to 2 additional classes:
`imblearn.over_sampling.SVMSMOTE` and
`imblearn.over_sampling.BorderlineSMOTE`.

There is also some changes regarding the API:
the parameter ``sampling_strategy`` has been introduced to replace the
``ratio`` parameter. In addition, the ``return_indices`` argument has been
deprecated and all samplers will exposed a ``sample_indices_`` whenever this is
possible.