Incremental Transition System Construction With Refinement For Region-Based Process Mining

Process discovery focuses on deriving process models from event logs to facilitate the analysis of complex workflows. Region-based approaches are effective in capturing concurrency and handling invisible transitions but suffer from high computational demands and reliance on predefined state information. This paper introduces an incremental algorithm for constructing \textit{transition systems} (TS) from event logs, integrating a refinement strategy to optimize state representation. By configuring state abstractions using past, future, or hybrid activity horizons, the proposed method significantly reduces transition system complexity. A refinement strategy is employed to determine optimal configurations for state representation, ensuring model quality through precision and simplicity metrics. The approach is evaluated using a case study, demonstrating its effectiveness in improving scalability and efficiency in region-based process discovery.