On behavioral process model similarity matching: A centroid-based approach

As business process models have a broad scope of applications, e.g., in science or in business administration, the problem of handling large amounts of process models arises. One helpful tool for dealing with this amount of models is to reduce it by using similarity measures in order to detect similar models that can be merged. A set of similar models may be replaced by one model. As a pure similarity of labels is often not enough to compare process models, other process perspectives are involved for calculating similarities. The current paper works on the process models' behavior, which is one such perspective. A problem that arises when comparing two models and that is covered in this paper is that one of a differing granularity of process steps. Due to this granularity problem M-to-N mappings are considered. The present paper provides a centroid-based and so easily computable method for calculating behavioral similarity values for process models, which is constructed for M-to-N mappings, and a short evaluation of it.