Behavior Differentiation of Process Variants With Invisible Tasks

Abstract

It is a fact that plenty of process variants are derived from the same base model in practical applications, and the main reasons include inevitable software maintenance and adaptability of process models. This fact raises a question that many process variants are with high similarity degree, and can not be differentiated from each other under the state-of-the-art similarity measurements. In order to differentiate similar-but-different process variants, we propose an approach of process variants’ behavior differentiation in this paper. The method analyzes the complex structures of process variants with invisible tasks, and utilizes task execution relationship to construct an integrated similarity measurement, which extends the existing similarity measurement capacity in terms of dealing with invisible tasks. It is proved that the proposed task execution relationship can capture the dominate features of similar-but-different process variant including invisible tasks, which means that as long as the task execution behaviors of two process variants are different, the corresponding behavior matrices must be different. Furthermore, a set of experiments are carried out, in order to evaluate the properties of effectiveness, semantic uniqueness expression and correctness of the proposed method. Meanwhile, the experimental results give evidences that the proposed method outperforms the existing model similarity measurements in accuracy.

Keywords

• Behavior differentiation
• behavior profiles
• change mining
• invisible task
• silent transition
• petri nets