Abstract: The consistency performance between the event log and the business process can be improved by model repair reasonably adjusting the deviation behaviors.The deviations observed the event log are repaired in a self-loop insert way,so the improvement of fitness will be prioritized over precision.To obtain the nonreplayed behavior mode between event log and process model,the reachable activity structure was divided into several fragments and the conformance was checked according to the behavior relationship.The optimal alignment produced by the behavior patterns during replay was used to detect the occurrences,locations and potential behavior relationships of deviations.Furthermore,the directly following deviation elements  were constructed into repairable substructure.Therefore,the precision could be improved by reducing the number of deviations in the event log.M-repair plugin was applied to evaluate different data sets,and the result showed that the precision was significantly promoted ensuring the fitness compared with the existing method.

Key words: model repair, reachable activity graph, optimal alignment, directly following deviations, deviation substructure

摘要： 模型修复通过对偏差行为的合理调整,改善事件日志与业务流程之间的一致性性能。采用自循环插入方式对日志中可观测的偏差活动进行模型修复,将优先考虑适合度的提升而忽视精度。为获得事件日志与过程模型之间不可回放的行为模式,根据行为关系将其可达活动图表划分为若干个片段并进行服从性校验。利用回放过程中行为模式所产生的最优对齐检测偏差的发生、位置以及潜在行为关系,将具有直接跟随关系的偏差元素构建为可修复的子结构,从而通过减少事件日志中的偏差个数而改善精度。通过实验使用M-repair插件在不同数据集上进行评估,结果表明该方法相较于现存方法在保证适合度的前提下可显著提升精度。

关键词: 模型修复, 可达活动图表, 最优对齐, 直接跟随活动, 偏差子结构