DISCOVERING PROCESS MODELS CONTAINING XOR BRANCHES

Abstract

This study aims to investigate the robustness of the process discovery algorithms implemented in the PM4Py library. We created synthetic event logs to serve as benchmark datasets for evaluating process discovery methods in terms of the complexity of the event logs. Specifically, we developed a test framework using process models containing XOR branches. For simple XOR branches, all the examined algorithms (Alpha Miner, Inductive Miner, and Heuristics Miner) effectively uncovered the underlying process models. However, for more complex scenarios with events occurring in varied structural positions, heuristic and inductive approaches proved to be more reliable.