Mining Constraints from Reference Process Models for Detecting Best-Practice Violations in Event Logs

TL;DR

This paper introduces a framework that automatically mines and applies constraints from reference process models to detect violations in event logs, reducing manual effort and improving conformance checking.

Contribution

It presents a novel method for extracting relevant best-practice constraints from model collections and applying them to real event logs for violation detection.

Findings

Effective detection of best-practice violations in real-world logs

Automated selection of relevant constraints from reference models

Demonstrated applicability on real process model collections

Abstract

Detecting undesired process behavior is one of the main tasks of process mining and various conformance-checking techniques have been developed to this end. These techniques typically require a normative process model as input, specifically designed for the processes to be analyzed. Such models are rarely available, though, and their creation involves considerable manual effort.However, reference process models serve as best-practice templates for organizational processes in a plethora of domains, containing valuable knowledge about general behavioral relations in well-engineered processes. These general models can thus mitigate the need for dedicated models by providing a basis to check for undesired behavior. Still, finding a perfectly matching reference model for a real-life event log is unrealistic because organizational needs can vary, despite similarities in process execution. Furthermore, event logs may encompass behavior related to different reference models, making traditional conformance checking impractical as it requires aligning process executions to individual models. To still use reference models for conformance checking, we propose a framework for mining declarative best-practice constraints from a reference model collection, automatically selecting constraints that are relevant for a given event log, and checking for best-practice violations. We demonstrate the capability of our framework to detect best-practice violations through an evaluation based on real-world process model collections and event logs.