[Paper Review] Process Mining for Python (PM4Py): Bridging the Gap Between Process- and Data Science

Alessandro Berti, Sebastiaan J. van Zelst+1 more|arXiv (Cornell University)|May 15, 2019

Business Process Modeling and Analysis117 citations

TL;DR

PM4Py is a Python library that enables customizable process mining algorithms and seamless integration with data science tools, addressing the gap between process mining and data science workflows.

ABSTRACT

Process mining, i.e., a sub-field of data science focusing on the analysis of event data generated during the execution of (business) processes, has seen a tremendous change over the past two decades. Starting off in the early 2000's, with limited to no tool support, nowadays, several software tools, i.e., both open-source, e.g., ProM and Apromore, and commercial, e.g., Disco, Celonis, ProcessGold, etc., exist. The commercial process mining tools provide limited support for implementing custom algorithms. Moreover, both commercial and open-source process mining tools are often only accessible through a graphical user interface, which hampers their usage in large-scale experimental settings. Initiatives such as RapidProM provide process mining support in the scientific workflow-based data science suite RapidMiner. However, these offer limited to no support for algorithmic customization. In the light of the aforementioned, in this paper, we present a novel process mining library, i.e. Process Mining for Python (PM4Py) that aims to bridge this gap, providing integration with state-of-the-art data science libraries, e.g., pandas, numpy, scipy and scikit-learn. We provide a global overview of the architecture and functionality of PM4Py, accompanied by some representative examples of its usage.

Motivation & Objective

• Lower the barrier for algorithmic development and customization in process mining compared to existing tools.
• Enable easy integration of process mining algorithms with other data science libraries in Python.
• Foster a collaborative ecosystem for sharing code and results in process mining.
• Provide thorough documentation and rigorous testing to ensure algorithmic stability.
• Support large-scale experiments and reproducible research in process mining.

Proposed method

• Use a strict separation of objects, algorithms, and visualizations across dedicated PM4Py packages.
• Provide factory methods as unified access points for algorithm variants (e.g., Alpha Miner classic and plus).
• Support multiple event data structures (event logs, event streams, pandas data frames) and conversions between formats.
• Implement core process mining techniques (discovery, conformance checking, enhancement) and quality measures (fitness, precision, generalization, simplicity).
• Offer visualization through Plot, GraphViz, NetworkX, and Pyvis integrations for process models and social networks.

Experimental results

Research questions

• RQ1 How can process mining algorithms be made easily extensible and customizable in Python?
• RQ2 How can PM4Py integrate process mining with existing Python data science libraries?
• RQ3 Can PM4Py support large-scale experimentation and collaboration within the process mining community?
• RQ4 What is the maturity and adoption level of PM4Py in academic and practical settings?

Key findings

• PM4Py provides Alpha Miner and Inductive Miner implementations and aligns with conformance checking and performance measurement.
• PM4Py supports event logs, event streams, and pandas data frames for scalable data handling.
• PM4Py uses factory methods to manage algorithm variants and ensure backward compatibility.
• PM4Py enables visualizations for Petri nets, directly-follows graphs, and social networks using GraphViz, NetworkX, and Pyvis.
• PM4Py 1.0 released in 2018 and PM4Py 1.1 in 2019, with growing usage in courses and integration with other projects.