Classification and evaluation of timed running schemas for workflow based on process mining

The system running logs of a workflow contain much information about the behavior and logical structure between activities. In this paper, a mining approach is proposed to discover the structural and temporal model for a workflow from its timed running logs. The mining results are represented in the formalized form of Petri nets extended with two timing factors that allows validation or verification the actual behaviors, especially the temporal constraints between activities. According to the reachability graph of the extended Petri net model mined, all running schemas of a workflow can be generated, which defines the temporal constraints between running activities. By calculating the earliest and latest start time of each activity, the earliest starting and latest existing time of each state in the running schema can be determined. Based on the temporal relations between the timing factors of each running state, the running schemas can be classified into six classes. The effects of the six classes of running schemas on the implementation of the whole workflow are evaluated so as to obtain the best one that can ensure the workflow is finished in the shortest time. The standards for the ideal, reliable and favorable running schemas and their existence conditions are discussed, which can be used to evaluate the running logs and control the future running of a workflow. © 2008 Elsevier Inc. All rights reserved.