Modeling the Integration of Machine Learning into Business Processes with BPMN

Machine Learning (ML) models offer diverse and wide-ranging capabilities to improve business processes. These can be an important success factor for increasing the degree of automation of process execution. Nevertheless, there is currently a lack of possibilities to explicitly model business processes which contain ML. This paper presents requirements for tools supporting explicit modeling of ML in business process models. In addition, it is elaborated how these aspects could be formally modeled because more formal modeling of ML cases allows more options for analyzing and using (e.g., executing) models. Overall, the presented approach describes a comprehensive tool concept for the step-by-step support of explicit modeling the integration of ML applications in business processes modeled with Business Process Model and Notation (BPMN), so that actual process models without ML can easily be extended to process models with ML. For this purpose, besides the extension of existing process patterns, a catalog of so-called sub-process templates is presented, which enables process engineers to derive application-specific sub-process models for different ML functions. Each sub-process template represents multiple BPMN process models and is expressed using the Case Management Model and Notation (CMMN) standard.