Repairing Process Models With Logical Concurrent and Casual Relations via Logical Petri Nets

Process mining techniques can build process models according to event logs generated by enterprise information systems. However, some event logs cannot be correctly replayed by the model, which needs to be repaired. When new activities are found in event logs, the existing models repair methods cannot accurately find the positions where to add them. By considering relations between new and original activities, this paper proposes an approach to repair logical Petri net-based models. We first define two new kinds of relations between new and original activities called logical concurrent and casual relations, respectively. Then, a ladder matrix is constructed, and the differences between an original process model and event logs can be obtained. Next, a deviation matrix is used to record these differences. New activities are added to construct a concurrent relation with original activities, or inserted behind the original activities to construct a casual relation. Through performing experiments on diagnosis data in a hospital, the correctness and effectiveness of the repaired model has higher fitness and precision than those by the previous approaches.