Model Repair Method Based On Logical Petri Nets And Alignment

In a rapidly evolving business environment,enterprises and organizations support their business processes through the use of information systems.A large number of event logs generated by system execution provide a foundation for the research in the field of process mining.Because process mining is extracting knowledge from event logs to discover,monitor,and improve actual processes.However,the advent of the digital age has led to the increasing number and rapid change of business processes,and it has allowed for differences between the inherent business processes and the actual execution of the behavior recorded by the system.From different dimensions and perspectives,process mining can be divided into three main fields:process discovery,consistency detection and repair.Process discovery techniques do not consider the initial model,but rather build a model that is best suited to the actual behavior being executed.Conformance detection techniques treat the original model as a specification and only diagnose and quantify deviations without building a new process model.The model repair not only considers the original model,but also extracts effective information from the event log to expand and improve the model.The effective information may be deviation location,deviation frequency,consistency detection value and so on.Therefore,model repair can also be regarded as an in-depth study based on consistency detection.However,existing studies on conformance detection,whether based on alignment or Token,generally have two problems:First,most of the studies regard all noncompliance behaviors as deviations,resulting in the inevitable occurrence of the optimal alignment at the same low cost;Second,in model repair,each deviation is repaired independently,without considering the overall situation.To sum up,this thesis replays the event log in the model according to the related theories of Petri net,and proposes the adjusted optimal alignment based on priority repair set and model repairing based the logical Petri nets.The main contents are as follows:(1)Propose extended alignment and optimal alignment based on move places.When calculating the alignment between a known process model and the actual event log,if there is a deviation(log move or model move),most of the existing methods cannot determine the exact position of the deviation in the process model,which is not conducive to the follow-up research.In this thesis,the concept of move places is introduced to supplement the alignment,so the concepts of supplemental alignment and optimal supplemental alignment are proposed,and the method of constructing supplemental alignment is given,so as to find and accurately locate the position of deviation in the model.(2)The way of selecting the optimal alignment at the same low cost based on the priority repair set is proposed.When there are multiple optimal alignments of equal low cost,the existing method is to select them by changing the parameters of the cost function,which can only reduce the number of optimal alignments to a certain extent.This thesis not only introduces the concept of priority repair sets,but also proposes the principle of maximizing the number of priority repair sets to obtain the multiple sets of optimal alignment,so as to select the adjustment optimal alignment.(3)In view of the fact that the deviation between the log and the model is log move,a log move repair method based on logical Petri nets and the adjustment optimal alignment is proposed.Existing repair methods are mostly based on traditional Petri nets.When repairing models with strong logical relationship between events,they either cannot be repaired or the model repaired is very complex.This thesis not only puts forward the concept of complementary sublog,at the same time,based on the two conditions of whether the activities in the supplementary sublog has been replayed and whether there is a concurrency structure in the replayed activities,a model repair method based on logical Petri nets and adjustment optimal alignment is proposed,digging out the generated logic transitions and its logic function.The restored logical Petri nets process model is simplified on the basis of ensuring the fitness and accuracy.Figure[11]Table[12]Reference[57]...