ReACMe: Repetition Aware Clustering Methodology for Business Process Log Collections

Caterina Luciani, Luigi Bucchicchio, Andrea Morichetta, Marco Piangerelli, Andrea Polini

Abstract: Process-Aware Information Systems (PAIS) are extensively employed to support organizational workflows, with configurations that often differ across various usage contexts. Analyzing the event logs they generate is essential for understanding this variability; however, traditional process mining techniques often face scalability challenges, particularly when dealing with loops and a large number of process instances. This paper introduces ReACMe, a parametric, unsupervised clustering methodology that bypasses model generation by leveraging n-gram-based features and a repetition-aware dissimilarity measure. Using the k-medoids algorithm, ReACMe effectively groups similar logs and allows to identify representative medoids. The approach is validated on both public datasets and a real-world e-government scenario, demonstrating its efficiency and practical applicability.

External IDs:doi:10.1007/978-3-032-04375-7_14