摘要
鉴于工作流在逻辑信息组织和协同工作方面的优势,近年来被广泛应用于各行各业。通过工作流技术实现的业务流程管理可以协调多种资源执行生产工作或服务,为客户产生价值。然而,在业务流程执行的过程中,可能会发生异常情况,阻止其按照预定的方式执行,给业务流程的目标带来风险,需要在流程执行的过程中提前预测发现异常,尽早做出调整。因此,提出一种基于机器学习方法的异常检测方法,通过挖掘流程执行的日志记录和活动执行时间信息,实时预测业务流程中的超期异常和流程行为异常。经过在公开数据集上的实验表明,所提算法能有效地找出潜在超期异常的流程,以及行为异常的流程。
Thanks to the advantages of workflow in logical information organization and collaborative work,it has been widely used in various industries in recent years.Business process,the most typical workflow,perform a series of activities to produce products or services and generate value for clients.However,In the execution of the business process,anomaly and exceptions may disturb the business process and pose risks for the process goals.Therefore,it is necessary to predict and find anomalies in advance during business process execution and make adjustments as soon as possible.For this reason,an anomaly detection method based on machine learning algorithms was proposed,which utilized the process execution logs and activity execution durations to predict the process time constraint violations and process behavior anomaly in real time.With experiments conducted on real-life business process logs,the method proposed could effectively identify the processes potentially violating time constraint and the processes with abnormal behavior.
引文
[1]RUSSELL N,VAN DER AALST W M P,TER HOFSTEDEA H M.Workflow exception patterns[C]//Proceedings of the International Conference on Advanced Information Systems Engineering.Berlin,Germany:Springer-Verlag,2006:288-302.
[2]ADAMS M,TER HOFSTEDE A H M,VAN DER AALSTW M P,et al.Dynamic,extensible and context-aware exception handling for workflows[C]//Proceedings of the International Conferences on the Move to Meaningful Internet Systems.Berlin,Germany:Springer-Verlag,2007:95-112.
[3]LI Hongchen,YANG Yun.Dynamic checking of temporal constraints for concurrent workflows[J].Electronic Commerce Research and Applications,2005,4(2):124-142.
[4]CHEN Jinjun,YANG Yun.Temporal dependency-based checkpoint selection for dynamic verification of temporal constraints in scientific workflow systems[J].ACM Transactions on Software Engineering and Methodology,2011,20(3):9.DOI:10.1145/1242060.1242063.
[5]EDER J,GRUBER W,PANAGOS E.Temporal modeling of workflows with conditional execution paths[C]//Proceedings of the International Conference on Database and Expert Systems Applications.Berlin,Germany:Springer-Verlag,2000:243-253.
[6]EDER J,PANAGOS E.Managing time in workflow systems[EB/OL].[2018-06-30].https://www.researchgate.net/profile/Johann_Eder/publication/2379826_Managing_Time_in_Workflow_Systems/links/00b49526a896002818000000/Managing-Time-in-Workflow-Systems.pdf.
[7]LEONG I F,SI Y W,BIUK-AGHAI R P.Predicting temporal exceptions in concurrent workflows[M]//Models for Capitalizing on Web Engineering Advancements:Trends and Discoveries.Hershey,Pa.,USA:IGI Global,2012:196-218.
[8]EDER J,PICHLER H.Duration histograms for workflow systems[C]//Proceedings of the IFIP TC8/WG8.1 Working Conference on Engineering Information Systems in the Internet Context.Berlin,Germany:Springer-Verlag,2002:239-253.
[9]SI Y,HOI K,BIUKAGHAI R P,et al.Run-based exception prediction for workflows[J].Journal of Systems and Software,2016,113:59-75.
[10]MISIC D,STOJKOVIC M,DOMAZET D,et al.Exception detection in business process management systems[J].Journal of Scientific&Industrial Research,2010,69(3):188-193.
[11]LUO H,LIU J,LIU X,et al.Predicting temporal violations for parallel business cloud workflows[J].Software-Practice and Experience,2018,48(4):775-795.
[12]DAU H A,CIESIELSKI V,SONG A,et al.Anomaly detection using replicator neural networks trained on examples of one class[C]//Proceedings of the Asia-Pacific Conference on Simulated Evolution and Learning.Berlin,Germany:Springer-Verlag,2014:311-322.
[13]LIU F T,TING K M,ZHOU Z H.Isolation forest[C]//Proceedings of the 8th IEEE International Conference on Data Mining.Washington,D.C.,USA:IEEE,2008:413-422.
[14]DJEDOVIC,A.Credit requirement event logs[D].Sarajevo,Bosnia and Herzegovina:University of Sarajevo,2017.