摘要
流程挖掘是从事件日志中自动构建流程模型,并利用其分析、增强、监测实际的业务流程.然而不完整的事件日志会严重影响流程挖掘的结果.因此,修复事件日志是提高过程挖掘结果准确性的举措之一.现有的修复事件日志技术主要修复事件日志中的缺失活动,很少考虑修复日志中多个缺失的属性.现实中日志除了活动缺失,还存在属性缺失现象.针对此问题,提出了一种基于Bert的神经网络模型,用来修复事件中多个属性的缺失.该方法从数据的角度出发,通过Bert模型的预训练任务学习事件中属性之间的依赖关系,根据属性的前后文信息预测缺失的属性值.最后使用公开可用的真实事件日志对所提出的方法进行实验评估,评估结果表明本文所提出方法可以修复事件日志中多属性的缺失,且验证了该方法的准确性.
Process mining automatically constructs process models from event logs and uses them to analyze,enhance,and monitor actual business processes.Reparation of event logs is one of the initiatives to improve the accuracy of process mining results because incomplete event logs seriously affect the results of process mining.Meanwhile,the current techniques for repairing event logs mainly repair missing activities in event logs but rarely consider repairing multiple missing attributes in logs.In reality,logs have missing attributes in addition to missing activities.To address this problem,this research proposes a Bert-based neural network model for repairing multiple missing attributes in events.The method learns the dependencies between attributes in events from a data perspective by per-training tasks of the Bert model,and predicts the missing attribute values based on previous and subsequent contextual information of the attributes.Finally,the researchers experimentally evaluated the accuracy of the proposed method by using publicly available real event logs,and the evaluation results show that the proposed method can repair the missing multiple attributes in event logs.
作者
张振虎
王丽丽
Zhang Zhenhu;Wang Lili(School of Mathematics and Big Data,Anhui University of Science and Technology,Huainan 232001,China)
出处
《洛阳师范学院学报》
2024年第2期17-22,共6页
Journal of Luoyang Normal University
基金
国家自然科学基金资助项目(61572035,61402011)
安徽省自然科学基金资助项目(2008085QD178)。
关键词
修复事件日志
缺失属性
Bert模型
预训练任务
多属性修复
reparation of event log
missing attribute
Bert model
per-training task
multi-attribute reparation