Abstract: There are a large number of events exist in event logs that contain activities of both order and originators,also named organizational dimension information.Control flow discovery algorithms can automatically construct the control flow process model from the event log,and the organization discovery algorithm builds a social network.If two dimensions can be combined and displayed in a model,a more complete process organization view can be provided,which is helpful for processing and organization analysis.Therefore,a DoubleDimensional Genetic Process Mining Method based on Executor Process Tree (BdSm) was proposed.Inductive miner was used to prepare a highquality initial population for genetic mining algorithm,which could generate high-quality control flow dimension process models.A measurement of distance between activities based on the similarity of executor was proposed.The process model of control flow was extended by organizational dimension information,and a double-dimensional process model based on the executor process tree was generated.The simulation logs and four open event logs were used to verify BdSm,and the results showed that BdSm could generate a process model of highly comprehensive quality in the control flow dimension.Meanwhile,typically working models and organizational structures could be found through organizational dimension information.

Key words: process mining, genetic mining algorithm, control flow dimension, organizational dimension, executor process tree

摘要： 事件日志记录数量众多的事件,不仅包含与活动控制流相关的内容,还记录有关活动执行者的信息,即组织维度信息。控制流发现算法从事件日志中自动构建控制流过程模型,组织维度发现算法则构建社交网络模型。如果能合并两种维度,在同一个模型中进行展示,则能够提供更完整的过程组织视图,有助于更准确地对过程以及组织进行分析。因此,提出一种基于执行者过程树的双维度遗传过程挖掘方法(BdSm)。一方面,使用Inductive Miner预挖掘以优化遗传挖掘算法初始种群,达到生成高质量的控制流模型的目的;另一方面提出日志中活动之间距离的度量方法,能有效度量活动在组织层面的相似度,同时使用执行者信息扩充控制流过程模型,基于执行者过程树生成双维度的过程模型。通过模拟日志以及4个公开事件日志集对所提方法进行验证,结果表明,在控制流维度,所提方法能够生成较高综合质量的过程模型,同时借助组织维度信息,还能够发现典型的工作模式及组织结构。

关键词: 过程挖掘, 遗传挖掘算法, 控制流维度, 组织维度, 执行者过程树