Exploiting General Purpose Big-Data Frameworks in Process Mining: The Case of Declarative Process Discovery

Declarative process models are a flexible tool to capture the process model in unstructured and highly variable business scenarios. Extracting declarative process constraints, however, is a computationally intensive task and, as the volume of log data increases, efficiently extracting constraint template occurrences becomes challenging. While there have been efforts to improve the performance of declarative constraint extraction, we argue that solutions can be seamlessly derived by adapting application-agnostic pattern analysis techniques for big data to this task. In this work, we propose a solution for the efficient extraction of declarative constraints, specifically those described in the Declare language, without the limitation of developing tailored systems for declarative processes. We build on top of a recent scalable framework, named SIESTA, which can perform efficient pattern analysis on large log files. Our approach yields promising results, significantly outperforming the existing Declare Miner and MINERful solutions.