Process mining builds a bridge between data mining and process modeling and analysis.Process mining is the extraction of effective information from the data and resources of the actual business process system.The process model can both describe the activities of the system and simulate the running sequence between activities.The behavior of process model is compared with that recorded in the event log to find their commonness and heterogeneity.When the process model and the event log do not match on the process,the process model is repaired to make the process model better reflect the actual business system and improve the performance of the model,so as to ensure that the information system maintains good consistency with the actual business process.Conformance checking and model repair of models containing concurrent structures are very important in process mining.Based on logic Petri nets,this paper proposes different model repair methods for cyclic concurrent structures,concurrent blocks,concurrent and sequential structures,incomplete concurrent and choice structures.(1)For cyclic concurrent structures,a model repair method based on the identity place set and the concurrent transition set is proposed.By adding the identity place set to extend the optimal extension alignment,a method of calculating the general or logic optimal extended alignment is proposed.The deviation is located by judging the relation among the identity place set of continuous log activities,the corresponding transition pre-set,and the concurrent transition set.And a dynamic model repair method is proposed based on the logic Petri nets.(2)For concurrent blocks,a model repair method based on the concurrent branch and the concurrent activity mapping set is proposed.The methods of computing the concurrent transition pair and choice transition pair are given based on process trees.The deviation is determined by judging the relation among the precursor and successor of log activities in the concurrent activity mapping set and the activities of the concurrent branch.And a model repair method is proposed based on the logic Petri nets.(3)For concurrent and sequential structures,a model repair method based on the model-tokan tuple and the trace-place sequence is proposed.All activities of event logs are forced to enable in the model,the method of computing the model-tokan tuple is given.According to the different deviation types of different structures,the different deviation positions are obtained combining the model-tokan tuple and trace-place sequence.And different model repair methods are proposed based on the logic Petri nets.(4)For incomplete concurrent and choice structures,a model repair method based on the choice relation set and concurrent relation set is proposed.The algorithms of computing the choice relation set and concurrent relation set are given based on process trees,including the choice activity set,concurrent activity set,branch set,etc.According to log activities and model activities in the optimal alignment,different deviation positions are obtained combining the activities of the choice relation set and concurrent relation set.According to different deviation positions,different model repair algorithms are proposed based on the logic Petri nets.This paper takes the business process of a hospital in Qingdao as the background example.Through the simulation experiment,the four different repair methods proposed in this paper based on logic Petri nets are compared with traditional repair methods,proving the correctness and effectiveness of our methods.Models repaired by our methods are more accurate and concise.The repaired models by our proposed methods improve precision and simplicity while maintaining a high degree of fitness. |