Extended Alignments Based Process Models And Logs And Model Repair Techiques

Model repair is a new type of process mining approach.The inputs of the approach are event logs and process models.By analyzing the event logs,the deviations between the event logs and the process models can be obtained.Then,the process model can be repaired.The approaches ensure that the repaired models are as close to the original models as possible.The thesis takes medical procedures in a hospital of Tsingtao as background.With the development of actual process,new activities occur in the medical procedures.The original model cannot accurately describe the real process,thus the model need to be repaired.The existing approaches focus on finding and repairing the deviations,and think less on the conformance of the repaired model.Since the repaired process model has some problems and shortcomings by existing approaches,major research is as follows.The extended alignments are proposed based reachable markings of Petri nets.And an approach to obtain extended alignments is proposed.For a trace and a process model,the alignments can obtain moves on log and moves on model.The moves illustrate the deviations between the trace and the process model.However,it is hard to confirm where the deviations appear in a process model by the alignment.The extended alignments not only find deviations,but also confirm location of deviations.For moves on log in extended alignments,an improved repairing approach is proposed.The subprocesses are added to the original model as loops by the existing approaches.It leads the repaired model has a low precision.The proposed approach collects the sublogs,and subprocesses mined by the sublogs are inserted into an original model rather than being added as loops.The repaired model has a higher precision.For the Petri nets with choice structures,a repairing approach based choice structures is proposed.However,the proposed approach before still allows some traces that do not belong to the event logs.Combined with process trees of Petri nets,the choice structure of Petri nets can be identified easily by finding the non-leaf nodes of process trees.Then,the choice branch deviations are defined,and an approach to obtain the choice branch deviations in an extended alignment is proposed.These choice branch deviations constitute the choice branch log.By the choice branch log,new choice branches are added to a proper choice structure of the model.Finally,the correctness and efficiency of the proposed approach are illustrated by some examples of the business processes of thoracic surgery in a hospital of Tsingtao.And the conformance of the repaired models by the proposed approach is better than other existing approaches.