Recommending Interesting Results in Process Mining Analysis

Abstract

Identifying relevant insights in the results of process mining is challenging and time-consuming since it requires considerable knowledge and relies heavily on manual efforts to find adequate subsets of data with the insights. Current tools lack proper support to help the user in this manual process. To mitigate this, some approaches have been developed to provide guidance based on clustering traces and finding differences in basic Process Performance Indicators using statistical tests. However, some insights might not only be based on differences and they might require looking into other subsets of traces. In this paper, we make a proposal to guide users to find results with interesting insights in process mining analysis. It receives an event log and provides subsets of traces with interesting insights along with possible explanations about why they are interesting. We achieve that by using data mining measures of interestingness in process metrics. We illustrate the potential use of our proposal with a real use case.