Abstract: Business process tends to change in real world application,which is called process drift.Traditional process discovery technique assumes process to be in a steady state and does not takes account of process drift.Process drift can be classified into sudden drift and gradual drift.Sudden drift has been well solved while gradual drift not.For this problem,an approach to detect gradual drift based on sudden drift detection was proposed,which could divide the event log into multiple sublogs.For any three successive sublogs,if the middle one's behavior was a combination of the other two's,then the middle one was generated by gradual drift.Experiments on synthetic logs showed that the proposed approach had better performance than state of the art.

Key words: process mining, process discovery, concept drift, change detection, event logs

摘要： 过程漂移是指过程模型在服务期间发生了变更或演化,分为突发漂移和渐变漂移两种类型。传统的过程发现技术假定业务过程是稳定不变的,忽略了过程漂移的存在。目前突发漂移已有较好的检测方法,但渐变漂移没有。以突发漂移检测为基础提出一种渐变漂移检测方法。该方法利用突发漂移点将日志划分为多个阶段,然后通过日志的行为语义混合判断某个阶段是否为渐变漂移。实验结果表明,所提算法的检测效果优于目前最好的算法。

关键词: 过程挖掘, 过程发现, 概念漂移, 变更检测, 事件日志