Towards multi-dimensional clustering of business process models using Latent Dirichlet Allocation

Tobias Dumont, Peter Fettke, Peter Loos

In: Tagungsband Multikonferenz Wirtschaftsinformatik. Multikonferenz Wirtschaftsinformatik (MKWI-16). Multikonferenz Wirtschaftsinformatik (MKWI-16) March 9-11 Illmenau Germany Seiten 69-80 1 Universität Illmenau 2016.


Large process model collections are hard to handle and deciders face the challenge of diverse and unstructured process landscapes. Classifications of business process models allow structuring the process landscape and therefore ease the decision making of managers. Current modeling tools support mainly hierarchical classifications although methods like tagging, evolving from Web 2.0, provide a more flexible systematization. Furthermore, they lack an automated classification of large process collections. In the paper at hand we present a methodology for an automated, tagging like, multi-dimensional-clustering of business process models. The processes are allocated to thematic clusters based on Latent Dirichlet Allocation which has formerly been used for text classification. We analyze strengths and weaknesses of the proposed method and present possible application scenarios. A prototypical implementation has been used to evaluate the method on the SAP reference models.

Weitere Links

ilm1-2016100012.pdf (pdf, 14 MB )

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence