PM4Py

PM4Py is a Python library providing state-of-the-art algorithms and tools for process mining.

It is developed and maintained by Process Intelligence Solutions GmbH (PIS).

Getting started

Installation instructions, tutorials, and the full documentation are maintained on the PM4Py website:

• Installation guide
• Tutorials and examples
• Feature overview
• API documentation

License

The open-source version of PM4Py, available on GitHub, is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0).

A separate version of PM4Py is available for commercial use in closed-source environments under a different license. See PM4Py Licensing for details.

Optional dependencies

The default installation covers mainstream usage. Additional integrations are grouped by feature in pyproject.toml:

Extra	Installs
calendars	workalendar
connectors	requests
llm	openai, requests
ml	pyemd, scikit-learn
ocel	jsonschema, pyarrow
polars	polars (dataframe support)
solvers	cvxopt
visualization	pyvis
windows	pygetwindow, pynput, pywin32 (Windows only)

The stable extra pins the exact dependency versions validated by the project and used in CI.

Polars

Polars 1.40.1 and 2.x are supported. Joins used in filtering and enrichment preserve event order, including when lazy queries run on the streaming engine (the default in Polars 2.0).

PuLP

SciPy is the default LP solver; PuLP is optional. If PuLP is installed, PM4Py selects the matching backend for PuLP 3.x or 4.x. PuLP 4 requires CBC to be installed separately, either with python -m pip install "pulp[cbc]" or by putting cbc on your PATH.

Release notes

See CHANGELOG.md for incremental updates.

Contributing

Before contributing, please review the contributing guidelines and Contributor License Agreement (CLA).

For a source checkout, install the project with python -m pip install -e .. Build, CI, and lint dependencies are standard dependency groups and can be installed with, for example, python -m pip install --group lint (requires pip 25.1 or later).

Third-party dependencies

As a scientific library in the Python ecosystem, PM4Py relies on external libraries. The licenses of all direct dependencies are in the third_party/ folder, and third_party/LICENSES_TRANSITIVE lists every transitive dependency with its license.

Citing PM4Py

If you use PM4Py in your scientific work, please cite:

Alessandro Berti, Sebastiaan van Zelst, and Daniel Schuster (2023). PM4Py: A process mining library for Python. Software Impacts, 17, 100556. https://doi.org/10.1016/j.simpa.2023.100556

DOI | Article

BibTeX:

@article{pm4py,
  title   = {{PM4Py}: A process mining library for {Python}},
  author  = {Berti, Alessandro and van Zelst, Sebastiaan and Schuster, Daniel},
  journal = {Software Impacts},
  volume  = {17},
  pages   = {100556},
  year    = {2023},
  issn    = {2665-9638},
  doi     = {10.1016/j.simpa.2023.100556},
  url     = {https://www.sciencedirect.com/science/article/pii/S2665963823000933}
}

Legal notice

This repository is managed by Process Intelligence Solutions GmbH. Further information about PIS is available at processintelligence.solutions.