.. _balance:

=====================
The balance submodule
=====================

The balance submodule provides methods for rectifying class imbalance by
either augmenting or down-sampling the dataset. As with most skoot methods,
each of these functions works on either Numpy array objects or Pandas
DataFrames.

|

As of version 0.19, there are three methods of balancing data:

* `Under-sampling (the majority class) <./generated/skoot.balance.over_sample_balance.html#skoot-balance-under-sample-balance>`_
* `Over-sampling (the minority class) <./generated/skoot.balance.over_sample_balance.html#skoot-balance-over-sample-balance>`_
* `SMOTE <./generated/skoot.balance.over_sample_balance.html#skoot-balance-smote-balance>`_

These balancing functions are *not* transformers, since balancing should never
be applied to test data. They are simply functions that should be applied to
training data prior to fitting a model.

|

See :ref:`balance_examples`

.. raw:: html

   <br/>