支持复杂结构的混成过程挖掘方法  

Hybrid process mining method supporting complex structures

作　　者：朱锐[1,2] 张志幸 莫启[1,2] 李彤[1,2] 马自飞[1] 黎彬 ZHU Rui 1,2 ,ZHANG Zhixing 3,MO Qi 1,2 ,LI Tong 1,2,MA Zifei 1,LI Bin 1(1. School of Software, Yunnan University, Kunming 650091, China;2. Key Laboratory in Software Engineering of Yunnan Province, Yunnan University, Kunming 650091, China;3.School of Software,Zhejiang University, Hangzhou 310058, Chin)

机构地区：[1]云南大学软件学院,云南昆明650091 [2]云南大学云南省软件工程重点实验室,云南昆明650091 [3]浙江大学软件学院,浙江杭州310058

出　　处：《计算机集成制造系统》2018年第7期1653-1670,共18页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(61662085);云南省教育厅科学研究基金资助项目(2017ZZX227);云南省教育厅科研重点资助项目(2015Z018);云南大学数据驱动的软件工程省科技创新团队项目(2017HC012);阿里活水计划资助~~

摘　　要：为解决传统过程挖掘算法在处理蕴含复杂结构的海量日志时的低效低质问题,提出一种支持复杂结构的混成过程挖掘方法。该方法首先将事件日志转化为具有发生次数的直接后继图,以支持活动间基本关系的判定;通过过程树对已发现的两两活动间的基本关系进行抽象与合并,进而对日志进行更新,反复迭代直到整个日志中的所有具有基本关系的活动被全部发现。若待发现模型由基本块组成,则挖掘结果为基于块的过程模型;若待发现模型包含复杂结构,则通过混成使用基于区域的方法对复杂结构进行发现。最终利用活动重构操作对挖掘结果中已抽象为过程树的部分进行细化,从而获得最终结果。为了进一步提升挖掘效率,还提出并行化的发现与重构方法。大量基于真实数据的实验结果表明,该方法的挖掘效率和挖掘精确度达到了较好的水平。To solve ineffective and low-quality in exiting process mining algorithm for dealing with massive amounts of complex construction logs, a novel hybrid process mining which was able to work in the presence of complex construction was proposed. The direct subsequent diagrams was extracted which had the number of occurrences from event logs to estimate the basic relationships in inter-activity. The basic relationships in the inter-activity by the process tree were abstracted and merged, and the logs were updated until all activities with basic relationship were discovered in all logs. If the yet to be discovered modules were made up from basic blocks, the block-based process model was employed to discover the complex construction; if the modules contain complex construction, the hybrid region-based approach was performed to discover the complex construction. The activity refactoring to refine the process tree was exploited which had been abstracted in mining results. Based on this representation, a parallel discovery and refactoring approach was further proposed to improve the efficiency of process mining. Our comprehensive evaluations showed that the approach could achieve highly accurate and efficient process mining over lots of real data.

关 键 词：过程挖掘 过程树 复杂结构 基于块的过程模型 活动重构 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]