Master Thesis Opportunity: Mining Reference Models from Large Business Process Collections

While there is some academic interest in the mining of reference models from process model collections, little is known about the feasibility of these approaches for practical purposes. Out of this purpose, the topic covers the identification of candidates for reference model mining in the wild.


Academic literature on reference model mining from process model collections is relatively sparse, and existing approaches typically focus on control flow properties, i.e., activity orderings. In practice, however, other model properties, such as diagram name, non-graphical diagram-level and element-level attributes, as well as non-control flow elements such as roles/resources (BPMN pools and lanes), documents, and data objects, are just as relevant. In particular, natural language properties are promising candidates for straightforward process variant identification across process landscapes, and roles, documents, and IT system definitions should not come as an afterthought in reference models. This master thesis sets out to explore these practical perspectives, given the academic state-of-the-art.

To this end, we will identify process model properties and algorithms for identifying reference model candidates in a large collection of business processes. Specifically, this thesis includes the following tasks:

(i) Identify relevant process model properties for clustering models into reference model- oriented groups.

(ii) Evaluate different clustering algorithms on a real-world dataset.

Find more information here.

If you are interested, please get in touch with Prof. Rehse.