Abstract: Determining the similarity degree between process models was very important for their management, reuse, and analysis. Current approaches either focused on process model's structural aspect, or had inefficiency or imprecision in behavioral similarity. Aiming at these problems, a novel similarity measure which extended an existing method named Transition Adjacent Relation (TAR) with improved precision and efficiency named TAR* was proposed. The ability of measuring similarity was extended by eliminating the duplicate tasks without impacting the behaviors. For precision, TARs was classified into repeatable and unrepeatable ones to identify whether a TAR was involved in a loop. Two new kinds of TARs were added, one related to the invisible tasks after the source place and before sink place, and the other representing implicit dependencies. For efficiency, all TARs based on unfolding instead of its reach ability graph of a labeled Petri net were calculated to avoid state space explosion. Experiments on artificial and real-world process models showed the effectiveness and efficiency of the proposed method.

Key words: transition adjacent relation, unfolding, Petri nets, behavioral similarity

摘要： 无中文摘要信息

关键词: 无中文关键词信息