Abstract: Process mining aims to discover behavioral relationships between process objects and their associated activities from business process event logs,offering technical support for the understanding,analysis,and optimization of business processes.However,traditional process mining approaches are typically based on a single case concept,making them less effective in dealing with the complex interactions involving multiple process objects.As business processes grow in complexity,object-centric process mining techniques have attracted increasing attention.Nonetheless,existing methods are insufficient in handling compliance checking for object-centric processes,often losing critical process interaction information,which results in inaccurate deviation detection and difficulty in identifying the sources of deviations.To address these issues,a token-based object-centric process compliance checking method was proposed,which constructed an object graph from object-centric event logs and extracted process executions from it.Subsequently,redundant cross-layer dependencies within the process executions were removed,and events were divided into tiers to obtain binding sequences.Token replay was then performed in a multi-object Petri net model according to the binding sequences,generating replay results and deviation outcomes.The method quantifies compliance through a defined fitness measure,while local diagnostic techniques were employed to highlight significant deviations in the process.Finally,a comparative analysis with traditional methods based on publicly available cases demonstrated that the proposed approach effectively addressed the limitations of traditional methods in accurately detecting deviations and identifying their sources,thereby providing robust support for the analysis of complex multi-object processes.

Key words: conformance checking, multi-object Petri nets, token replay, process execution, process mining

摘要： 流程挖掘旨在从业务流程事件日志中发现流程对象及其关联活动间的行为关系,为业务流程的理解、分析和优化提供技术支持。然而,传统流程挖掘方法基于单一案例概念,难以有效处理涉及多个流程对象的复杂交互行为。随着业务流程复杂性的增加,对象为中心流程挖掘技术逐渐受到关注。但已有方法在处理对象为中心流程合规性检查方面存在不足,易丢失流程交互信息,导致偏差识别不准确且难以确认偏差来源。为解决上述问题,本文提出一种基于托肯重演的对象为中心流程合规性检查方法。该方法从对象为中心事件日志中构造对象图,并从中提取流程执行。然后,删除流程执行中的跨层冗余依赖,并按梯队划分事件获取绑定序列。随后,在多对象Petri网模型中按照绑定序列进行托肯重演,得到重演结果和偏差结果。通过定义的拟合度量化合规性,并辅以局部诊断技术,凸显流程中的严重偏差。最后,基于公开案例与传统方法进行对比分析,验证本文方法可以有效解决传统方法无法准确识别偏差以及难以确认偏差来源的问题,为多对象复杂流程分析提供支撑。

关键词: 合规性检查, 多对象Petri网, 托肯重演, 流程执行, 流程挖掘