Precision-Guided Minimization of Arbitrary Declarative Process Models

Declarative model minimization is a computationally expensive task. State-of-the-art approximation techniques rely on hard-coded heuristic functions based on properties of constraint templates, which requires rework when new templates are added and cannot handle models expressed as arbitrary logical formulas. We present a precision-based heuristic function that requires no pre-configuration and can handle arbitrary constraints, provided they can be mapped to a finite automaton. The approach is evaluated on real-world datasets, where it outperforms state-of-the-art methods while accepting a wider range of inputs.