/feature_engine

Feature engineering package with sklearn like functionality

Primary LanguagePythonBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

Feature Engine

Python 3.6 Python 3.7 Python 3.8 License CircleCI Documentation Status

Feature-engine is a Python library with multiple transformers to engineer features for use in machine learning models. Feature-engine's transformers follow scikit-learn's functionality with fit() and transform() methods to first learn the transforming parameters from data and then transform the data.

Feature-engine features in the following resources:

Blogs about Feature-engine:

Documentation

En Español:

More resources will be added as they appear online!

Current Feature-engine's transformers include functionality for:

  • Missing Data Imputation
  • Categorical Variable Encoding
  • Outlier Capping or Removal
  • Discretisation
  • Numerical Variable Transformation
  • Variable Creation
  • Variable Selection
  • Scikit-learn Wrappers

Imputing Methods

  • MeanMedianImputer
  • RandomSampleImputer
  • EndTailImputer
  • AddMissingIndicator
  • CategoricalImputer
  • ArbitraryNumberImputer
  • DropMissingData

Encoding Methods

  • OneHotEncoder
  • OrdinalEncoder
  • CountFrequencyEncoder
  • MeanEncoder
  • WoEEncoder
  • PRatioEncoder
  • RareLabelEncoder
  • DecisionTreeEncoder

Outlier Handling methods

  • Winsorizer
  • ArbitraryOutlierCapper
  • OutlierTrimmer

Discretisation methods

  • EqualFrequencyDiscretiser
  • EqualWidthDiscretiser
  • DecisionTreeDiscretiser
  • ArbitraryDiscreriser

Variable Transformation methods

  • LogTransformer
  • ReciprocalTransformer
  • PowerTransformer
  • BoxCoxTransformer
  • YeoJohnsonTransformer

Scikit-learn Wrapper:

  • SklearnTransformerWrapper

Variable Combinations:

  • MathematicalCombination
  • CombineWithReferenceFeature

Feature Selection:

  • DropFeatures
  • DropConstantFeatures
  • DropDuplicateFeatures
  • DropCorrelatedFeatures
  • SmartCorrelationSelection
  • ShuffleFeaturesSelector
  • SelectBySingleFeaturePerformance
  • SelectByTargetMeanPerformance
  • RecursiveFeatureElimination
  • RecursiveFeatureAddition

Installing

From PyPI using pip:

pip install feature_engine

From Anaconda:

conda install -c conda-forge feature_engine

Or simply clone it:

git clone https://github.com/solegalli/feature_engine.git

Usage

>>> import pandas as pd
>>> from feature_engine.encoding import RareLabelEncoder

>>> data = {'var_A': ['A'] * 10 + ['B'] * 10 + ['C'] * 2 + ['D'] * 1}
>>> data = pd.DataFrame(data)
>>> data['var_A'].value_counts()
Out[1]:
A    10
B    10
C     2
D     1
Name: var_A, dtype: int64
>>> rare_encoder = RareLabelEncoder(tol=0.10, n_categories=3)
>>> data_encoded = rare_encoder.fit_transform(data)
>>> data_encoded['var_A'].value_counts()
Out[2]:
A       10
B       10
Rare     3
Name: var_A, dtype: int64

See more usage examples in the Jupyter Notebooks in the example folder of this repository, or in the documentation.

Contributing

Details about how to contribute can be found in the Contributing Page

In short:

Local Setup Steps

  • Fork the repo
  • Clone your fork into your local computer: git clone https://github.com/<YOURUSERNAME>/feature_engine.git
  • cd into the repo cd feature_engine
  • Install as a developer: pip install -e .
  • Create and activate a virtual environment with any tool of choice
  • Install the dependencies as explained in the Contributing Page
  • Create a feature branch with a meaningful name for your feature: git checkout -b myfeaturebranch
  • Develop your feature, tests and documentation
  • Make sure the tests pass
  • Make a PR

Thank you!!

Opening Pull Requests

PR's are welcome! Please make sure the CI tests pass on your branch.

Tests

We prefer tox. In your environment:

  • Run pip install tox
  • cd into the root directory of the repo: cd feature_engine
  • Run tox

If the tests pass, the code is functional.

You can also run the tests in your environment (without tox). For guidelines on how to do so, check the Contributing Page.

Documentation

Feature-engine documentation is built using Sphinx and is hosted on Read the Docs.

To build the documentation make sure you have the dependencies installed. From the root directory: pip install -r docs/requirements.txt.

Now you can build the docs: sphinx-build -b html docs build

License

BSD 3-Clause

References

Many of the engineering and encoding functionalities are inspired by this series of articles from the 2009 KDD Competition.