Research On Process Model Repair Based On Optimal Alignment

Taking the event log as the starting point for process mining,it is usually necessary to preprocess the log data,including data cleaning and data repair.Consistency detection is used to measure the consistency between the event log and the actual business process execution.For consistency detection,locating the location of the deviation and considering the cost of rediscovering the business process is relatively large.In contrast,model repair is more practical.However,for the preprocessing of log data,the existing research work,on the one hand,filters the log according to the frequency,on the other hand,repairs the existing event log,but most of them consider the event log as a simple activity sequence,and the repair work also focuses on simple event reordering.However,in fact,the log data also records time stamps,resources,and other attributes.For incomplete event logs with missing events,we simply consider the rearrangement of event sequences and completely ignore the impact of conflicts between attributes on log repair.The common consistency detection method based on optimal alignment determines the optimal alignment according to the principle of minimum alignment cost,so as to determine the deviation position,ignoring the behavior reflected by some short logs.The accuracy of the model based on this repair needs to be further improved.After determining the location of the deviation,the existing research on improving the efficiency of model repair starts from the whole in the way of single thread,but there is less research on model repair combined with the idea of block structure partition.Based on the work done by predecessors,the main research contents of this paper are as follows:(1)The repair of the event log does not simply consider the rearrangement of events,and the event log also contains other attributes.This paper mainly focuses on the repair of time stamps in the event log.If the event sequence is determined solely based on the maximum confidence value,it will face the problem of small confidence value due to missing events.Therefore,it is proposed to determine the sequence based on the maximum average confidence value,and further consider whether there is conflict between attributes between events,The average event attribute confidence algorithm is proposed.After determining the sequence order,the event timestamp is estimated to repair it.(2)Consistency detection technology aims to measure the degree of consistency between the log and the model behavior.At present,the position of deviation is determined based on the optimal alignment.There is a problem of ignoring the behavior reflected by some short log traces because of considering the minimum alignment cost,which affects the effect of consistency detection.In this paper,a consistency detection method based on optimal alignment is proposed.All paths in the model are extracted in the preprocessing stage.According to the relative optimal alignment algorithm(ROA),the log sequence and path sequence are compared one by one.The alignment with less log moving cost is selected as the relative optimal alignment,and the functions of calculating the fitting degree and accuracy are given.Finally,the actual case analysis is carried out to repair the position of the deviation determined based on the optimal alignment and the optimal alignment respectively.The results show that the accuracy of the repaired model based on the former is higher than that of the latter.(3)Most of the existing model repair methods are single thread repair from the global perspective,and less consideration is given to the correspondence between the detected deviation position and the area to be repaired.In this paper,the sub group behavior relationship is considered to repair the model,the logs are filtered according to the necessary event order,and the logs and the model are decomposed into sub group logs and model sub blocks.Extract the sub logs that do not fit in the sub group logs based on the optimal alignment,and analyze the behavior relationship between the sub log activities.For the behavior relationship of adding and deleting,insert and delete activities and corresponding arcs and place in the sub block.For the changed behavior relationship,change the corresponding arcs in the sub block to support the new behavior.Figure [seventeen] Table [seventeen] Reference [ninety]...