A Foundational Approach for Managing Process Variability

Abstract

Abstract. A business process often shows different variations in a large organisation, due to different legal requirements in different countries, deviations in the IT infrastructure, or organisational differences. These variants are documented in separate independent process models. Management of these variants imposes various challenges. Invariant behaviour needs to be identified and redundancies among the variants have to be avoided. In this paper, we address these questions by defining a set-algebra for behavioural profiles. These profiles represent a behavioural abstraction of process models that can be computed efficiently. We trace back many questions of process variability management to set-theoretic operations and relations defined for behavioural profiles. As a validation, we apply our approach to an industry model collection.

A Short Survey on Process Model Similarity

“…Hence, abstractions have been proposed that only consider the order in which two activities can be executed in any process instance [12]. Two variants are (1) the transition adjacency relation [13,14] that consider pairs of activities that can be executed directly after each other (non-transitive) and (2) weak order relations [4,15] which consider any pair of activities that can be executed after each other eventually (transitive). An extension of the latter are behavioural profiles, which distinguish these relations by mutual exclusion, strict, and interleaving order [16,17].…”

Multilevel business process modeling

“…Previous work has identified flexibility [19,18] and variability [29] as important issues in business process modeling. For example, a health insurance claim requires that a medical examination takes place during the assessment phase, unlike other insurance categories.…”

Generating Hierarchical Reference Process Model Using Fragments Clustering

“…However, these modeling approaches are not easy to achieve when individual process variants have already been created for different specific purposes independently [7]. In this context, a major challenge is that how to automatically build a reference model inductively based on the existing process variants instead of building the model manually by domain experts.…”