Volatile environments force companies to adapt their processes, leading to so called concept drifts during run-time. Concept drifts do not only affect the control flow, but also process data. We differentiate between internal and external data. Internal data refers to the attributes of the process tasks such as resource assignments. External data stems from process-external sources such as sensors. An example are manufacturing processes where a multitude of machining parameters are necessary to drive the production. This paper provides online algorithms for both, concept drift detection for internal data and finding the reason for concept drifts based on external data, employing the concept of process histories. The feasibility of the algorithms is shown based on a prototypical implementation and the analysis of a real-world data set from the manufacturing domain.

Top