On the impact of low-quality activity labels on data-driven process analytics

While event log data quality is recognized as a crucial concern in process mining, the impact of event log errors on different types of process mining tasks has remained largely unexplored. This paper aims to fill such a gap by analyzing how errors affect analysis results. We focus specifically on errors affecting the activity labels, which are crucial for virtually any process mining task, as they capture the steps carried out in a process. Our objective is to assess whether and to what extent different types of errors that impact the accuracy of activity labels affect the performance of the three typical process mining tasks: process discovery, conformance checking, and predictive process monitoring. We discuss a set of experiments in which errors on activity labels in both synthetic and real-life logs are injected using a simulator of the well-known event log data quality imperfection patterns. The results of the experiments are used to extract insights into the design of data cleaning pipelines for different process mining tasks.