Study On Process Variants Mining And Merging Based On Event Logs

Over the past few years,many large organizations have used process-aware information systems to manage their processes,which in the course of operation have resulted in organizations having multiple branches.Because of the need to encompass the needs of different organizational managers and comply with different standards,the business processes developed by these branches have similar and differentiated business objectives,that is,process variations.The need for the branch to adapt flexibly to changing needs makes the process variations it produces difficult to maintain.At present,research on process variants mainly focuses on two aspects:(1)How to find process variants from logs becomes a key problem to solve multi-version business process mining;(2)How to identify the common characteristics of existing business flows to eliminate process redundancy is also of great practical value.In order to solve the above two problems,this paper starts from the behavioral relationship between process model activities and the structural features of process model,so as to enrich the research content of process variants and expand the prospect of application.In order to solve the mining problem of multi-version business process,the benchmark log is extracted from the original event log according to the strict order relation in the behavior contour,so as to mine the context tree of the event log.Then,the benchmark log and the context tree are taken as inputs,and the benchmark log is classified into several parts by the clustering method;Finally,the clustered results are mapped to the original event log to mine configurable process variations.By PROM simulation,the experimental results show that the proposed method has good adaptability.To quickly build a single process model that meets new requirements,it is often necessary to merge multiple existing process variants to eliminate process redundancy.First,the matching scores of different combinations among a plurality of process variations are calculated to select a pair of process variation combinations with the highest matching scores.Then,according to our proposed merging algorithm,the common part of the process variant combination is extracted to create the corresponding copy,and the difference between the variants is appended as the branch of the configurable connector to work,and then a single configurable process model is generated.The combined process model can capture all the behaviors of the input model and trace the nodes in the model to determine which process variant the nodes are from.However,the process variation merging method based on Petri net can not merge multiple process variations,and can not detect the unexpected behavior in the merging model.Therefore,the process variant is divided into a single variant fragment using the grouping fairness in the Petri net;Secondly,the control flow path of the variant fragment is extracted and its matrix representation is constructed based on it,then the process variant is merged into a single process model.Finally,by demonstrating that the merged process model is displayed,all behaviors of the input process model can be captured,and undesired behaviors can be prevented from being generated in the merged model.Based on the behavioral relationships of process model activities and the structural features of process model,this paper discusses two methods of process variant mining and process variant merging,the main contents are as follows:(1)A method of mining process variants directly from the event log is proposed,especially when there is no reference model in the business process.(2)The validity of the algorithm is verified by using the event logs of the bank credit process and the library loan and return book process.(3)A Petri net-based process variants merging method is proposed to merge existing process variants to obtain a single process model.(4)An improved process variant merge method is proposed,which is mainly used to solve two problems in process variant merge method:Cannot disposable merge multiple process models and unexpected behavior detection,so that the merge algorithm is more perfect.Figure 14 table 23 reference 110...