关键词: 业务流程, 模型结构, 日志行为, 相似度计算

Abstract: To address the issues of existing business process similarity computation methods,the process similarity based on process logs was computed,these existing methods were improved based on the characteristics of process logs as input and two improved methods were proposed.One method reconstructed the process model through the log and computed the process similarity based on the model structure.However,to embody the tendency of some execution path or branch in the process log,the log trajectory was replayed on the process model,the best reacting model was used to find the matching log track by A* algorithm,and the executed flow relationship was identified under the best reacting model.On this basis,the weighted BPMN model generated by flow relationship which was weighted was used to compute the process similarity.The other method did not need to reconstruct the model and compute the process similarity based on the log behavior sequence.To support the task similarity computing,the distance between tasks was defined.Thus,the distance of task sequence supporting task distance was obtained,and the business process similarity was computed by using the multiple set of task sequence.Through the experiments on different real data sets,the similarities and differences of the two methods in similarity computing were analyzed,and the scope of application of the two methods was summarized.

Key words: business process, model structure, log behavior, similarity computing