面向长轨迹的高效业务流程合规性检查方法OA

Conformance checking aims to identify discrepancies and similarities between business process event log behaviors and model behaviors.Alignment,the most effective technique for conformance checking,precisely locates deviations between log and model behaviors.However,as event logs grow in scale and complexity,particularly when logs contain activities outside the model' s activity set,existing alignment methods face performance bottlenecks when processing long traces.This is mainly because model-external activities cause the alignment search space to expand dramatically,resulting in severe computational inefficiency.To address this issue,this paper proposes a long trace oriented alignment(LTA)method to improve alignment calculation efficiency and thereby enhance the overall performance of conformance checking.The core idea of the proposed method is to pre-filter activities that do not belong to the process model in order to reduce the alignment search space.Specifically,the complete activity set is first extracted from the process model,and the original log is filtered based on this set to generate a new log that contains only model-related activities.The filtered log is then aligned with the process model to obtain an incomplete alignment sequence.Finally,the complete optimal alignment result is reconstructed using the index information of the filtered activities.The proposed method has been implemented in PM4Py(process mining for Python)and evaluated using six groups of publicly available event logs and process models.Experimental results show that,compared with the traditional A*-based alignment method,the proposed approach achieves an average improvement of 41.3%in alignment efficiency while maintaining 100%conformance accuracy.