摘要
鉴于现有量化模型行为的研究大多数集中在结构化的过程模型层面,实际挖掘得到的业务模型多为非结构化,提出对非结构化过程模型的轨迹计数方法。整合现有结构化转换方法,最大化地计算非结构化过程模型的结构化后轨迹数,并对带同步条件的半结构化过程模型设计轨迹计数方法;对部分不能结构化的过程模型,基于完全前缀展开设计了对非自由选择结构的轨迹计数算法,包括循环和非循环两种情况;将两种方法与块结构过程模型的轨迹计数方法结合。在人工合成模型和实际的过程模型数据集上,对所提算法进行有效性测试和性能评估,实验结果表明,所提方法在其规定的作用范围内能够准确量化非结构化过程模型的行为。
In the conformance checking of the field of process mining, how to quantify the behavior of the process model and then judge the quality of the model is a hot research topic. Most of the existing researches on quantitative model behavior focus on the structured process model level, while the business models obtained by actual mining are mostly unstructured. Therefore, a trace counting method for unstructured process models was proposed. By integrating the existing structuring methods, the numbers of unique paths after structuring the unstructured process model were calculated, and the unique paths counting method for the semi-structured process model with synchronization conditions was designed. Then, for some of the process models that could not be structured, the unique path scounting algorithm for a non-free choice structure based on complete prefix unfolding was designed, including both cyclic and acyclic cases. Then, the two methods were combined with the unique paths counting method in a block-structured process model. On the artificial synthesis model and actual process model data sets, the validity test and performance evaluation of the proposed algorithm were carried out. The experimental results showed that the proposed method could accurately quantify the behavior of unstructured process models within a specified range.
作者
邓雅方
林雷蕾
闻立杰
刘瑷玮
钱忱
王建民
DENG Yafang;LIN Leilei;WEN Lijie;LIU Aiwei;QIAN Chen;WANG Jianmin(School of Software,Tsinghua University,Beijing 100084,China;School of Management,Capital Normal University,Beijing 100048,China)
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2022年第10期3100-3114,共15页
Computer Integrated Manufacturing Systems
基金
国家重点研发计划资助项目(2019YFB1704003)
国家自然科学基金资助项目(62021002)。
关键词
过程模型
轨迹长度
轨迹数
完全前缀展开
模型复杂性
process model
trace length
number-of-traces
complete prefix unfolding
model complexity
