Abstract: Business process similarity computation is an indispensable task of process management.A novel approach was proposed to compute the process similarity more accurately by considering both structure and behavior characteristics.Process models were converted to Business Process Graphs (BPG).The weight to each edge in the BPG was added according to the event log to obtain a weighted BPG.The weighted BPG that integrated the behavior of both process model and event log could provide more reliable and comprehensive similarity measure.The edit distance measure of two weighed BPGs was defined to compute the similarity.By comparison with existing approaches,the effectiveness of the proposed approach was demonstrated,and all approaches had been implemented in the open-source process mining toolkit ProM.

Key words: business process, process model structure, log behavior, weighted graph edit distance, similarity computing

摘要： 流程相似度计算是业务流程管理不可缺少的任务,为了更准确地计算流程之间的相似性,本文提出一种将模型结构与日志行为综合考虑的流程相似度计算方法,先将流程模型结构转化为业务流程图,根据日志中的行为信息对业务流程图的有向边进行加权,进而构建加权业务流程图。加权业务流程图同时包含了流程模型和流程日志中的行为信息,弥补了单一基于流程模型计算流程相似度时的不足。在此基础上给出了加权业务流程图编辑距离的定义,从而得到一种新的流程相似度计算方法。通过实验与已有基于流程模型结构和模型行为相似度计算方法进行对比,证明了所提方法的有效性。所提方法已在开源流程挖掘平台ProM工具中实现。

关键词: 业务流程, 模型结构, 日志行为, 加权图编辑距离, 相似度计算