Generating Block-Structured Parallel Process Models by Demonstration

Abstract

Programming by demonstration (PBD) is a technique which allows end users to create, modify, accommodate, and expand programs by demonstrating what the program is supposed to do. Although the ideal of common-purpose programming by demonstration or by examples has been rejected as practically unrealistic, this approach has found its application and shown potentials when limited to specific narrow domains and ranges of applications. In this paper, the original method of applying the principles of programming by demonstration in the area of process mining (PM) to interactive construction of block-structured parallel business processes models is presented. A technique and tool that enable interactive process mining and incremental discovery of process models have been described in this paper. The idea is based on the following principle: using a demonstrational user interface, a user demonstrates scenarios of execution of parallel business process activities, and the system gives a generalized model process specification. A modified process mining technique with the α|| algorithm applied on weakly complete event logs is used for creating parallel business process models using demonstration.

Weakly Complete Event Logs in Process Mining

“…Another improvement that we have achieved with weakly complete event logs compared to causally complete event logs is the property of weakly complete event logs that based on them, the models of parallel business processes can be interactively (using demonstration) generated, which is presented in detail in our paper [7].…”

“…In order to find and infer relations which were not performed (inferred relations), the system uses the event log footprint 1 and certain rules described later in this paper. The results and ways of realization of this idea are presented in our paper [7].…”

“…T ⊆ A, A is a set of activities, T is a finite set of transitions in the Petri net[4], σ is a trace such that σ ∈ L [2, 3] 7. The concepts of marking and tokens • are well known for Petri nets and precise definitions can be found in[2,3].…”