Study On Log Induced Process Model Repair

With the rapid development of the information age,enterprises are using processaware information systems to support their business process.In order to help enterprises quickly understand the business process,the original manual modeling method is abandoned,and the research on process mining technology is emerging.Process mining technology includes process discovery,conformance checking and process enhancement.Process discovery builds the event logs recorded by the information system into a process model.Conformance checking detects the deviation between the event log and the original process model.Process enhancement fixes the model that deviates from the actual business process.Due to the constant change of data in the information age,the data recorded in the enterprise information system is inconsistent with the process model in real life.In order to avoid bringing huge economic losses to the enterprise,the process model manager needs to adjust the existing process model in time.In order to reduce the cost of repair,this paper studies on the basis of the event log and the original process model,finds the differences between them,and then repairs the original model,so that the repaired model can not only replay the event log,but also be as similar as possible to the original model.This article considers two repair scenarios: online and offline.For offline scenarios,this paper proposes deviation detection and model repair based on process tree.This article assigns labels to each node in the process tree,and then aligns each trace in the event log with the process tree model.The deviation detected by the above method,namely the visible leaf node model move and log move is repaired accordingly.For the model move of visible leaf nodes,hidden transition is added to the corresponding structure in the process tree model to complete the repair.For log move,this paper finds the intermediate node closest to the visible leaf node,and exchanges the positions of the two in the corresponding optimal alignment,and then finds the deviation sub-log contained in the intermediate node in all the updated optimal alignment,and repairs the original model by comparing the behavior profile between the process tree and the sub-log.For online scenarios,this paper proposes a new method of deviation detection and model repair based on hidden Markov model.This method estimates the state of each observation unit in the event flow,and judges the deviation region by comparing the injection distance between the current state and the previous state.Then,the model is repaired by hidden transition.In order to better repair the original model,when a new activity appears in the case,according to the direct strict order relationship between the new activity and other activities,static repair is carried out based on logical Petri nets to determine the position of the new activity in the original model.Then,the above deviation field detection and repair method is used.This method can not only detect the deviation in real time,but also repair the detected deviation in time to avoid major mistakes.Figure [36] Table [4] Reference [70]...