Definition and detection of control-flow anti-patterns in process models

The correctness of process model is getting crucial as more and more process models begin to execute on workflow engines. Anti-pattern, as an effective and objective mean to verify the correctness of process model, is gaining increasing attention. Great loss could be avoided if control-flow anti-patterns were detected in an early time. In this paper, a control-flow anti-pattern detection approach for process models is proposed, which can support user-defined control-flow anti-pattern detection with different process modeling languages. To achieve this, firstly the control-flow structure of process model is transformed to refined process structure tree (RPST for short) via the cycle equivalence algorithm. Then CAPDL, a control-flow anti-patterns description language, is designed and an anti-pattern detection algorithm based on CAPDL is proposed. Finally, an anti-pattern detection experiment was conducted based on 278 real-world BPMN process models, the detection result of which shows that the proposed approach can detect user-defined control-flow anti-patterns effectively.