A Business Process Mining Method Based On Petri Net Filtering Technology

In recent years,business process management systems have attracted more and more attention from enterprises.A large amount of log data is generated during the enterprise operation management process.According to the log data,a business process model can be built to understand the running status of the system,which is convenient for management personnel to perform fault detection and system performance optimization.However,event logs often contain high frequency chaotic activities,low frequency chaotic activities,and infrequent behavior.The application of filtering technology in business process mining can filter the chaotic activities in business processes to simplify the flow relationship of business processes,avoid business process tedious,improve the suitability of mining business process models,and retain effective infrequency activities to optimize business processes.Therefore,this paper proposes a business process mining method based on Petri net filtering technology with certain theoretical significance and practical value.Most of the previous filtering techniques are modeled by high-frequency priority decision-making.The high frequency of event logs is considered to be the main behavior of business processes,so high-frequency chaotic activities are retained as main activities in business processes.Event logs with a small number of occurrences are considered to be directly filtered out by noise,and effective infrequency activity is filtered out as noise.Obviously this filtering method has certain limitations.In order to better deal with the above problems,this paper proposes a business process mining method based on Petri net filtering technology.The effectiveness of the proposed filtering algorithm is verified by examples.The main contributions of this paper are as follows:(1)The focus of process mining is to find frequent behaviors,while infrequent behaviors are often considered to be outliers or noise is ignored,but infrequent behavior may also be important for the management of business processes.Aiming at this problem,this paper proposes a mining method based on log automata and communication behavior profile for conditional infrequent behavior.First,the low-frequency execution log of the model is calculated by using the log automata to calculate the infrequent behavior of the low-frequency arc deletion log to become an abnormal automaton.and then processed,Then add attributed to the processed log.Finally,the conditional dependent values of communication characteristics between different modules are calculated to determine whether the conditional infrequent log is deleted or not,and the optimized event log is used to optimize the communication model.(2)Business process and can occur at any part of the business process.There is currently little research on chaotic activity in the filtered event log.Chaotic activities can be divided into high frequency chaotic activities and low frequency chaotic activities.For high frequency chaotic activities,the difficulty lies in how to separate high frequency chaotic activities from the main activities in the business process.In order to solve this problem,this paper proposes a method based on log automata and entropy filter log chaos activity.Firstly,the suspicious chaotic activity set with large entropy is calculated according to the direct pre-set rate and the direct post-set rate. Then,the log automata is constructed from the event log,and the activity set of the infrequent arc The chaotic activity set is obtained by calculating the intersection of the activity set of the infrequent arc is calculated from the log automaton model and the activity set with the large entropy value.and the active set of the entropy value in the log is taken to obtain the chaotic activity set.Finally,the dependent relation between the chaotic activity and other activities is determined by using the conditional occurrence probability and behavior profile,so as to decide whether to delete the chaotic activity completely in the log or to delete the activity in other positions by keeping the correct position of the chaotic activity in the log.An example shows that the filtering method can filter high frequency chaotic activities in the log.(3)For low frequency chaotic activities,the difficulty lies in how to distinguish low frequency chaotic activity from effective infrequent behavior.In order to solve this problem,a method of filtering chaotic activity based on entropy and behavior tightness is proposed.Firstly,according to the dependence between activities,the entropy is calculated to obtain the suspicious low-frequency chaotic activity set.Then the query model is constructed for the suspicious chaotic activity set.Finally,the tightness score of the query model and the business process is calculated according to the concept of behavior tightness.The low-level chaotic activity with low tightness score is directly filtered in the event log.Figure[27]Table[22]Reference[94]...