BPMN-Sim: A multilevel structural similarity technique for BPMN process models

Abstract

A better treatment of subprocesses by process model similarity techniques can provide improved results to stakeholders, allowing them to define which levels in the subprocess hierarchy are most relevant for similarity measurement. This article proposes a structural similarity technique – named BPMN-Sim – designed specifically for Business Process Model and Notation (BPMN), which considers multiple subprocess levels through a weighting factor in addition to supporting all simple and complex notation elements of BPMN 2.0.2. We focus on structural similarity as it is more sensitive to the element types and syntax rules that differentiate BPMN from other, simpler process modeling notations. We extended a generic structural similarity technique that relies on static vectorial signatures due to its expandability and adaptability to larger contexts and its potential for better efficiency. Due to the large number of elements in the BPMN notation, we developed a dynamic vectorial signature to avoid sparseness and inefficiency. Finally, BPMN-Sim results in values ranging from 0 to 1, which gives stakeholders good abstraction and comparability, and hence offers better meaningfulness, unlike other techniques whose results can range from 0 to a huge number. An empirical validation has shown a strong positive correlation between BPMN practitioner evaluations and results automatically computed by the proposed technique.