Abstract: As the most popular media for large-scale data,relational databases record amount of event logs during the operation of information system.However,traditional process mining techniques mainly deal with event logs in XES format,and every mining task needs to manually export the latest log file from the database.The whole process is very tedious,and it cannot make full use of the powerful data processing capacity of relational database.For these problems,the process mining strategy and algorithm for relational log were studied.To improve the process mining efficiency for relational logs,an approximate linear mining algorithm with high universal was proposed based on the fast sorting capability of relational database for large-scale event logs stored in relational databases,which had less invasive to the business database.The algorithm had been implemented on ProM platform,and the experiment based on large-scale event log showed the high efficiency of the proposed method.

Key words: process mining, directly-follow relation, relational database, experimental comparison, ProM platform

摘要： 关系数据库作为企业管理数据的主要工具,在信息系统运行过程中记录下大量事件日志。传统的流程挖掘技术主要处理用文件存储的XES格式日志数据,每次挖掘任务都需要手工从数据库导出最新日志文件,整个过程操作十分繁琐,且无法充分利用关系数据库强大的数据处理能力。针对该问题,研究了面向关系型日志数据的流程挖掘策略与算法。针对关系数据库中储存的大规模事件日志,利用关系数据库的快速排序能力,提出一种挖掘流程任务之间紧邻关系的近似线性挖掘算法,提高了关系型事件日志的流程挖掘效率。该算法对业务数据库侵入性小,具有较好的通用性。该算法已在开源软件平台ProM上实现,通过基于大规模事件日志的对比实验验证了该方法的高效性。

关键词: 流程挖掘, 紧邻关系, 关系数据库, 实验对比, ProM平台