基于可预测适合度的选择性模型修复

Alternative Model Repair Based on the Predictable Fitness

由于信息系统记录的行为不断变化,因此事件日志与给定模型之间往往存在偏差.事件日志可能产生2种不同类型的偏差,且每种偏差在偏差总数中的占比是不确定的.已有方法采用固定方式修复日志中非迭代偏差和自循环产生的迭代偏差,或在理想适合度被设定为1的前提下选择执行不同的修复方式,因而很难保证适合度与精度始终在合理范围.针对这一问题,提出一种修复方法可根据迭代可观测偏差总成本预测配置优化后的适合度,并在其满足给定阈值的情况下对所有偏差进行整体配置.当预测适合度不满足给定阈值时,进一步通过最优对齐发现事件日志与过程模型之间的变体,并根据每个变体的实际情况使用配置优化或者自循环插入的方式修复可观测偏差.仿真实验中对不同数据集进行了验证,结果表明:在始终保证适合度合理的前提下所提出方法能够最大程度地改善精度.

The recorded behavior in information system is constantly changing,so the deviation occurs between the event log and the given process model.Two kinds of deviations are produced by the event log,and the percentage of each deviation is uncertain in the total number of deviations.The iterative deviation generated by self-cycling and the non-iterative deviation in the event log is repaired by the same form in existing method.Furthermore,different repair form is alternatively executed under the ideal fitness of 1.Therefore,it is difficult to always ensure the fitness and precision in a reasonable range.To solve this problem,a new repair method is proposed to predict the configured fitness based on the total cost of the iterative observable deviations.Moreover,all the deviations are wholly configured when the predictable fitness meets a given threshold.The variant between the event log and the process model is found by optimal alignment when the predictable fitness does not meet the given threshold.Configuration optimization or form of self-loop insert is used to repair iterative observable deviation in the each variant based on actual condition.Simulation experiment is used to verify different data sets.The results show that the proposed method is beneficial to maximally improve the precision under ensuring reasonable fitness.