PM4Py-GPU: a High-Performance General-Purpose Library for Process Mining

TL;DR

PM4Py-GPU is a high-performance, open-source process mining library leveraging GPU acceleration to significantly improve analysis speed on large event datasets, addressing performance limitations of existing tools.

Contribution

The paper introduces PM4Py-GPU, a GPU-accelerated process mining library that enhances computational performance compared to traditional CPU-based tools.

Findings

Achieves significant speed-up in process mining tasks

Utilizes NVIDIA RAPIDS framework for parallel processing

Enables analysis of larger datasets efficiently

Abstract

Open-source process mining provides many algorithms for the analysis of event data which could be used to analyze mainstream processes (e.g., O2C, P2P, CRM). However, compared to commercial tools, they lack the performance and struggle to analyze large amounts of data. This paper presents PM4Py-GPU, a Python process mining library based on the NVIDIA RAPIDS framework. Thanks to the dataframe columnar storage and the high level of parallelism, a significant speed-up is achieved on classic process mining computations and processing activities.