基于活动的行为关系与关联时间的 多视角迹聚类方法.

Most of the commonly used trace clustering methods use a relatively single standard, such as using the activity sequence relationship, while ignoring the activity behavior relationship, time or resource attributes, which is unfavorable for some flexibly configured business process systems, as it hard to improve the quality of process mining. In order to solve such problems, this paper proposed a multi-perspective trace clustering method based on activity behavior relationship and association time. Firstly, this method constructed the control flow code according to the behavior relationship between activities. At the same time, in terms of time attributes, it used a group of nearest association activity pairs and their time differences to represent traces. Secondly, it used weighted aggregation to integrate the trace similarity under the two perspectives, and then adjusted the clustering results. Finally, the paper applied this method in the login system scenario and compared with other clustering methods on five real logs. The experimental results show that the method can find process scenarios from complex login systems, and verify the superiority of the method from three metrics of fitness, precision and Fl score. [ABSTRACT FROM AUTHOR]