Abstract: Due to the variability and dynamics of business processes in practical applications,the conformance between logs and given models needs to be checked and analyzed.The current conformance method aligns the log with the model directly to detect the overall deviation,while the deviation from the global observation will affect the accuracy of conformance checking.Therefore,a new method for analyzing conformance by searching for optimal alignment between logs and a given model was proposed.The selective subsequence reference activities was set to divide the type of alignment,and the minimum deviation alignment in the process of elimination was found through the arithmetic of alignment search.Then the optimal alignment was determined by analyzing the precision and behavior relations of the same deviation replay with different forms in the model sub-modules.The method was evaluated through case studies and comparative experiments of actual business process,and the results showed that accurate detection deviation could significantly improve the measurement value of fitness.By analyzing the detailed data of fitness measurement,the conclusion that narrowing the range of alignment comparison could reduce the calculated amount was obtained.

Key words: conformance checking, fitness measurement, minimum deviation, optimal alignment, deviation detection

摘要： 由于业务流程在实际应用中的多变性及动态性,需要对日志与给定模型之间的服从性进行校验与分析。目前的服从性校验方法主要将日志与模型进行对齐处理以检测整体偏差,但这种全局观测所获得的偏差会影响服从性校验的准确性,因此提出一种新方法对日志和给定模型间的最优对齐进行搜索来分析服从性。首先设定选择子序列划分对齐的种类,并通过对齐搜索算法在不断排除的过程中寻找最小偏差对齐;然后分析同一偏差以不同形式回放于模型子模块中的精度及行为关系确定最优对齐。该方法通过实际业务流程的研究案例及比较实验进行评估,其结果表明准确检测偏差能够明显提高适合度的测量值,并通过分析适合度测量的详细数据得出缩小对齐比较范围可减少计算量的结论。

关键词: 服从性校验, 适合度测量, 最小偏差, 最优对齐, 偏差检测