A Morphological Fragment Business Process Model Mining Method Based On Log Automata

In the process of business process analysis,business process management plays an increasingly important role and plays a crucial role in various process mining fields.A good business process model can maintain the normal operation of the enterprise system and improve the production efficiency of the enterprise.The purpose of process mining is to extract executable process knowledge from event logs and to monitor and improve real processes.Therefore,process mining technology has great practical application significance in business process development.At present,most of the process mining techniques adopt the principle of high frequency first,and filter the low frequency behaviors in logs directly.But some process systems include some infrequent behavior(eg:the escape system of a spacecraft,etc.)that occurs infrequently but critically in the system.Therefore,this paper proposes a morphological fragment business process model mining method based on log automata.Firstly,the behavior contour relation and quasi-indirect dependence relation among the transition activities are found out,and the hide transition activities in the process model are mined.Secondly,the sequence of event log is divided by the filtering operation technology of process cutting,which filters the noise activities in the event log and retains the infrequent behaviors that may contain valid information.Finally,the paper adopts the morphological fragment mining method and use the log automaton to calculate the infrequent arc of the activities in the event log,so as to filter the noise activities in the infrequent behaviors more accurately and further improve the accuracy of the process model.The main work of this paper includes following:(1)Aiming at the problem of hide transition in business processes,existing methods have some defects in the rationality of model mining and the mining of incomplete event logs.This paper proposes a hide transition method for business process mining with quasi-indirect dependency.Using integral linear programming to construct a dependency table among log activities,the constraint bodies among log sequences were found.Using the quasi-indirect dependency table to find the quasi-indirect relationship transitions that meet the requirements,and mining the hide transition in the quasi-indirect relationship transitions is beneficial to improve the rationality and appropriateness of the model.(2)For the infrequent behavior of the event log recorded by the business process,the existing research method directly filters according to the frequency of the log.This method causes the error to delete part of the effective low-frequency event log,and reduces the accuracy and consistency of the process model.This paper proposes a process cut method to filter noise in log activity.Process cuts not only take into account frequent behavior in log activity,but also behavior in low frequency mode.For the ring structure,the abnormal ring structure will cause anomalies in the edge structure of the flow chart,and the process cut can deal with the structure well.The filtered log is beneficial to improve the validity of the model to some extent.(3)For the optimization of construction model,a method based on log automata for morphological fragment mining process model is proposed.First,the event log sequence is converted into a log automaton model,and the log automata is used to calculate and verify the arc in the event log,and the unreasonable arc filtering process is performed.Then,according to the morphological fragment method,the event log is modularized,and the associated modules are found,and the same activity transitions in the associated activities are merged,and this step is iterated to obtain a complete process model.This method is effective in filtering infrequent behaviors,and is very effective in processing the process model of multiple sets of event logs,so that the process model is further optimized.Figure[31]table[32]reference[118]...