基于机器学习的流程异常预测方法

Anomaly prediction approach in business process based on machine learning

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.