计算机应用研究2025，Vol.42Issue(11)：3299-3306,8.DOI:10.19734/j.issn.1001-3695.2025.04.0116
面向直接后继关系交互演化的过程模型预测方法
Process model forecasting method for interactional evolution of directly-following relations
摘要
Abstract
PPM serves as a key task in process mining,aiming to predict future process behavior based on the current event log.Most existing PPM approaches primarily perform short-term predictions for individual process instances,such as next activi-ty prediction and remaining time estimation.These approaches offer limited prediction scopes and fail to provide a global per-spective on process evolution,making it difficult to reveal long-term trends in process model changes.To address this limita-tion,this paper proposed a PMF method based on time series analysis,which introduced a way to forecast the long-term evolu-tion of process models.The method transformed raw event logs into multivariate time series,systematically capturing the tempo-ral frequency evolution of all activity pairs(i.e.,directly-following relations)in the process.By modeling the mutual influ-ences among direct successor relationships,the approach predicted the future direct follower graph,thereby enabling long-range forecasting of the entire process model.Experimental results demonstrate superior prediction accuracy and stability compared to traditional time series approaches across multiple real-life event logs.These results indicate strong potential for practical appli-cations in monitoring and optimizing process behavior over time.关键词
过程模型预测/直接后继关系/时间序列分析/流程演变/过程挖掘Key words
process model forecasting/directly follows relation/time series analysis/process evolution/process mining分类
信息技术与安全科学引用本文复制引用
张润涛,方贤文..面向直接后继关系交互演化的过程模型预测方法[J].计算机应用研究,2025,42(11):3299-3306,8.基金项目
国家自然科学基金资助项目(61572035) （61572035）
安徽省重点研究与开发计划资助项目(2022a05020005) （2022a05020005）
安徽省自然科学基金资助项目(水科学联合基金)(2308085US11) （水科学联合基金）
