基于多实例Petri网的跨组织外包业务过程挖掘关键技术研究

This research project aims to develop novel process mining techniques that are capable of handling event logs generated from cross-organization outsourcing scenario to support the understanding, improvement, and redesign of real-life business processes. Although process mining techniques have been successfully applied in various domains, there still exists a lot of urgent challenges when considering the multi-instance sub-process pattern discovery in the cross-organization out-source scenario. More specifically, we plan to tackle these challenges from the following three perspectives: (1) We will introduce multi-instance Petri net and its correctness verification techniques to provide a solid theoretical foundation for formalizing, modeling and verifying cross-organization out-sourcing business processes; (2) To support the multi-instance sub-process pattern, a novel process discovery approach will be developed. Based the new approach, we can guarantee that the discovered cross-organization outsourcing business process models are correct by definition; and (3) We will develop new conformance checking metrics to quantify the quality of the discovery models with respect to the input event logs. The research output of this project can be used to provide theoretical models, discovery techniques, and quality evaluation metrics, and finally, support the redesign and improvement of cross-organization outsourcing business process models.

本项旨在开发能够处理跨组织业务外包场景下产生事件日志的新型过程挖掘方法，为企业业务过程理解、改进和重构提供依据。虽然过程挖掘技术已经在很多领域取得了成功应用，但是针对跨组织外包业务过程所涉及到的多实例子过程模式的挖掘仍存在许多亟待解决的挑战。具体而言，本项目将从以下三个方面的研究内容进行突破：(1)提出多实例Petri网模型与正确性判定方法，为跨组织外包业务过程形式化建模与验证提供理论基础；(2)开发新的过程挖掘方法，支持跨组织外包业务过程中多实例子过程模式的挖掘，并且保证挖掘跨组织外包过程模型的正确性；(3) 建立新的合规性度量方法，对跨组织外包业务过程模型进行质量评估。项目研究成果将会为跨组织外包业务过程提供理论模型、挖掘方法和质量评估方法，进而为跨组织外包业务过程模型重构和改进提供理论依据和技术支持。

本项目以跨组织外包业务场景下产生事件日志的过程挖掘算法研制为核心，经过三年的研发工作，圆满完成预期任务。共发表/录用论文49篇(其中，SCI论文29篇、EI论文19 篇)，申请/授权发明专利30项(其中，授权发明专利5项，1项完成成果转化)，授权软件著作权6项、出版英文学术专著1部，项目组成员累计参加国内国外学术会议18次，并做分组报告13次，相关科研成果在过程挖掘领域开源平台ProM和PM4PY进行发布。.具体而言，项目中取得的研究成果汇总如下：(1)提出了多实例Petri网模型与正确性判定方法，为跨组织外包业务过程形式化建模与验证提供理论基础；(2)开发了新的过程挖掘方法，支持跨组织外包业务过程中多实例子过程模式的挖掘，并且保证挖掘跨组织外包过程模型的正确性；(3) 建立了新的合规性度量方法，对跨组织外包业务过程模型进行质量评估；(4) 针对已有挖掘算法在处理大规模事件日志时存在的性能瓶颈，提出了一系列高效的事件日志采样算法；(5) 针对当前业务过程剩余时间预测性能和准确性等局限，提出了一种基于双向准循环神经网络和注意力机制的流程剩余时间预测模型来提高预测的性能和准确性；(6) 针对复杂时间日志的动态异质性，提出了业务过程事件日志的漂移检测与评估方法。