Khiops Python Library

This is the repository of the Khiops Python Library for the Khiops AutoML suite.

Description

Khiops is a robust AutoML suite for constructing supervised models (classifiers, regressors and encoders) and unsupervised models (coclusterings). With this package you can use Khiops via Python in two ways:

• with the module khiops.core: To use Khiops in its native way (Khiops dictionary files + tabular data files as input)
• with the module khiops.sklearn: To use Khiops with Scikit-Learn estimator objects (Pandas dataframes or NumPy arrays as input)

Installation

Using conda

conda create -n khiops-env
conda activate khiops-env
conda install -c conda-forge khiops

Using pip under Linux (in a bash shell)

python -m venv khiops-venv
source khiops-venv/bin/activate
pip install -U khiops

Using pip under Windows (in a powershell shell)

python -m venv khiops-venv
khiops-venv\Scripts\activate
pip install -U khiops

Other installation methods are documented at the Khiops website.

Requirements

• Python (>=3.10)
• Pandas (>=2.3.3)
• Scikit-Learn (>=1.7.2)

Documentation

The API Docs for the Khiops Python library are available here. Other documentation (algorithms, installation etc.) can be found on the Khiops site.

The library itself is documented with docstrings: for example, to obtain help on the KhiopsClassifier estimator and on the train_predictor function, respectively, you can use:

from khiops.sklearn import KhiopsClassifier
help(KhiopsClassifier)

from khiops import core as kh
help(kh.train_predictor)

License

The Khiops Python library is distributed under the BSD 3-Clause-clear License, the text of which is available at https://spdx.org/licenses/BSD-3-Clause-Clear.html or see the LICENSE.md for more details.

Credits

The Khiops Python library is currently developed at Orange Innovation by the Khiops Team: khiops.team@orange.com .