Abstract: To solve the problem that single firing sequence could not use the mature mining approach due to the absence of case information,a dual-view approach included model level and instance level was adopt.From the model perspective,the cycle was existed in a trace to ensure the correctness of mining foundation,and a concurrence block set to resolve the promiscuous problem caused by concurrent activities was proposed;from the trace perspective,a heuristic approach modified to accommodate cases division was presented to measure concurrent relationship so as to reduce the impact of noise,and a dependency table contained the concurrent relationship was built to separate the cases.By integrating all aspects,a framework to separate cases concurrently according to activities was put forward,and a best case based on fitness was selected.The real world data was used for extensive experiments sets to show the effectiveness and correctness of proposed method on single firing sequence mining.

Key words: process mining, single firing sequence, heuristic approach, case separating, Petri nets

摘要： 为解决因缺少挖掘所需案例属性的支持而无法使用成熟挖掘算法对单触发序列进行挖掘的问题,从模型层和实例层双视角进行研究,从模型角度证明轨迹中存在循环以保证挖掘基础的正确性,提出构建并发块集来解决并发活动对案例划分引起混淆的问题;从轨迹角度对启发式方法进行改进以适应案例划分,提出启发式的并发关系度量方法以降低噪声对并发关系挖掘的影响,通过构建含有并发关系的依赖关系表对案例进行划分。综合提出一个针对活动集并行化地进行案例划分,并根据其拟合度择优选择最佳案例的方法框架。通过大量基于真实数据集的实验展示了该方法针对单触发序列挖掘的有效性和正确性。

关键词: 过程挖掘, 单触发序列, 启发式方法, 案例划分, Petri网