Business processes have to constantly adapt in order to react to changes induced by, e.g., new regulations or customer needs resulting in so called concept drifts. By now techniques to detect concept drifts are applied on process execution logs ex post, i.e., after the process is finished. However, detecting concept drifts during run-time bears many benefits such as instant reaction to the concept drift. Introducing pro-cess histories as a novel way to detect and represent incremental, sudden, recurring, and gradual concept drifts through mining the evolution of a process model based on an event stream will face this challenge. There-fore, a formal definition of process histories is given, the concept of pro-cess histories is prototypically implemented and compared with existing approaches based on a synthetic event log.

Top