Research On Log Induction Change Mining Methods

With the rapid development of business process management and the updating and changing of business requirements,model change analysis and mining have gradually become an important component of model management.In the actual organization or company management process,the organizer has to deal with various complex business process changes.Therefore,the change mining problem of process model becomes an important part of business process management.The process change mining studies mainly focuses on the model-based research of its change region or change propagation problem.However,in the case where the business system is unknown or the actual process reference model is not given,how to mining the change of process behavior is an urgent problem to be solved in the field of change mining.Therefore,process behavior change mining based on event log has become a hot topic in the field of process management.At present,the research on model change analysis mainly focuses on three aspects.The first is based on the variation domain analysis of the process model behavior profile relation.The second is based on the problem of multi-view variation propagation of data flows and control flows.The last focuses on the study of process variants,mining variable fragments for business process design and so on.The above research mainly rely on the situation that the system model is pre-defined,and there are relatively few researches on process change only based on event logs.In order to make up for the limitations of the existing research scope and extends the existing methods,from the log point of view of event logs,the impact of log behavior changes on the model changes under the two conditions of complete logs and incomplete logs are taken into consideration.At the same time,the business process changes are further analyzed from the perspective of log implicit dependencies and change fragments.This paper mainly explores the change analysis of the process model based on event log based on the log behavior profile relation.The main research contents are as follows:(1)A process change location method based on the feasible trace of Petri net is proposed.Firstly,the event traces of the real process model and the feasible traces after the change are given.The minimal successor relation of the two groups of feasible traces is calculated separately.The calculated transitions successor relation and the original event traces are used to find out the inconsistent transitions in the successor relation,and the behavior relation with the change is analyzed to find the corresponding change.(2)A process change mining method based on incomplete log is proposed.Existing process change mining methods are either for the case of the process model is known or for the complete log mining analysis.However,the mining of behavior relation changes under incomplete log conditions has not been studied in depth.This method is analyzed from two aspects.Firstly,a motivation example is presented.From the perspecitves of completeness and incompleteness of the logs,the changes of the system behavior relation that may be caused by the single activity change operation are discussed.Secondly the log conjoint occurrence relation is used to calculate and analyze the process behavior change,and the simulation experiment is carried out by using ProM.(3)An implicit dependency change mining method based on log communication behavior profile is proposed.It mainly explores the implicit dependencies between transitions from the perspective of implicit places.Further improve the existing research method that focuses on the transitions.In the case that the logs are not perform the change operations,such as delete,insert,move.The process change which contains the implicit dependency is mined based on the log communication behavior profile relation and the transition successor relation.(4)A log-based process clustering method is proposed.Firstly,the low frequency events in the logs are filtered,and the common high frequency fragments are extracted by the log morphology fragments.The extracted common high frequency fragments are converted to cluster centers of similar logs by a formal automaton.Secondly,business combination method is proposed based on morphological fragments to generate frequent execution paths of process models.The similar equivalent morphological fragments are combined into a business,and the combined Petri net model is the clustering center of the process cluster.Figure[28] table[25] reference[102].