Abstract: To improve the retrieval efficiency of models,a similarity algorithm named TAR++ was presented based on importance of Transition Adjacent Relations (TARs).The main idea of TAR++ was to describe the transitions relationship through adding an importance argument on TARs,and present the similarity of models with the similarity of TARs sets.The experiment proved that the distance of TAR++ algorithm was satisfied the properties of distance metrics.The relative tests on actual business process of three corporations of SAP,East pot and BeiChe showed that the proposed algorithm had better efficiency,flexibility and operability.

Key words: unfolding net, behavioral similarity, transition adjacency relation, similarity algorithm evaluation

摘要： 为了提高模型的检索效率,提出一种基于变迁紧邻关系重要性的相似性算法TAR++,其主要思想是在流程中两两任务之间的紧邻关系上增加一个重要性系数,并将流程的相似度用带重要性的变迁紧邻关系集合的相似度表示。经证明,该算法对应的TAR++距离满足距离度量性质。在SAP、东锅、北车三个公司实际业务流程数据上进行的相关实验表明,该算法具有比较好的时间效率以及良好的灵活性和可操作性。

关键词: 展开网, 行为相似性, 变迁紧邻关系, 相似性算法评估