Capturing Ordered Flow in a Constraint-Based Formalism

Abstract

Many organizations use business process management to manage and model their processes. For the purpose of modeling business processes, two different types of modeling formalisms can be distinguished: Flow-based formalisms, resulting in procedural process models and constraint-based formalisms, resulting in declarative process models. Process models are considered procedural when the focus is on the order of activities. The execution scenario is explicit and designed within implicit business constraints. In contrast, declarative process models are recognized by explicit business constraints and the execution scenario is inferred from these business constraints. Several limitations of procedural models have been discovered in recent literature, which can be dealt with using a constraint-based formalism. Flow-based formalisms such as BPMN are considered the current standard for business process modeling. Consequently, a large amount of time and effort has already been spent by organizations on modeling their processes with a flow-based formalism. As such, the goal of this research is to develop a methodical way of transforming procedural process models into a constraint-based formalism. Furthermore, to alleviate the limitations induced by procedural formalisms, the process model concepts are interpreted in order to explicate the original business constraints it was based on. To realize this, six techniques were selected: Graph transformation, Well-formed BPMN transformation, related cluster pair similarity measurement, semantic process model similarity measurement, lexical analysis of activity labels and a Naïve Bayes classifier. These techniques were adjusted to be compatible with BPMN as source formalism and the Declarative Process Modeling Notation (DPMN) as target formalism. The techniques were tested using a sample set containing 103 BPMN process models. Resulting from the techniques was a set of transformation and interpretation rules, which were used to create a transformation method. The method was used to transform and interpret a different sample set, which contains 10 BPMN process models. An evaluation on the result of these transformations was used to improve and fine-tune the method. The method is able to transform 9 out of 10 process models. The 9 successfully transformed process models were reviewed by DPMN modeling experts and received an average grade of 5.6 (on a scale from 1 to 10) for utilizing the declarative properties in an advantageous manner. Based on the results of this review, final improvements were made to the method. In future research, additions can be made to the method. The method can be expanded with multiple source and target formalisms. Additional techniques can be added as well. For example, graph mining, graph edit distance, process mining, process behavior and BPMN-Q are considered as possible additions to the set of techniques.