Data-Driven OCL Invariant Patterns-Based Process Model Exploration for Process Mining

Process mining has become an effective and efficient means to manage organisations’ processes. A commonly used type of process mining today is conformance checking, which compares event log data with a given process model to detect possible conformance issues. Process specifications often contain data-driven constraints that the event log data must satisfy. However, the process model often fails to capture such restrictions adequately. Moreover, existing rule-based conformance checking techniques have commonly relied on rules that deal with execution orders or jointly executed conditions of activities within an event log but have yet to consider data-oriented constraints. In this paper, we introduce data-driven OCL (Object Constraint Language) invariant patterns that contain placeholders to facilitate the representation of such constraints from the process model. Within our approach, event log data is expressed as snapshots, and then we validate the snapshots with OCL constraints generated from our patterns. The proposed framework is experimented with and evaluated on a process belonging to an e-commerce domain.