Decomposed Mining Method And Application For Business Process Model Based On The Net Of Behavior Feature

Process mining is a technology for restoring the business process model based on the event log.Due to the diverse needs of the market,the interaction between models through multiple behavioral interactions under the operation of the system has become a new trend.However,the interaction of behavior produces a more complexity event log which increases the difficulty of process mining.Therefore,it has a great theoretical and practical significance to modularized the decomposed mining of the business process model.With the advent of the era of big data,the data of logs from the operation of system is increasingly huge and the algorithm of process mining based on event logs began to face challenges.For example,the explosive growth in the volume of information of logs makes the computational complexity of process mining has increased greatly and the operation efficiency of models has a linear decrease.The existing algorithms of process model merely aim to the event log which contains fewer features,activities and has a simple model requires.To the logs under the background of big data,this paper presents a mining method of interaction process model based on feature net.Firstly,analyzes the order relation of the internal characteristic among characteristics which corresponded by activities in logs,then mining the initial module nets preferentially;Secondly,according to the definition of the interface transition and feature net,traversal logs to mining interface transitions and add interface places for it;Lastly,using the view of the synthetic net,combine interactive modules into a perfect process model through interface places.The mining method can not only deal with logs that contain large number of features,but also mining the process model which interacted infrequent effectively.The other main contributions of this article are as follows:(1)For decomposing model based on simple event logs during the mining of business process model,this paper propose a mining method of module net based on the interface transition of Petri net.The method breaks the limitation of existing algorithms of decomposed miming that must mining the complete net preferentially through the analysis of the predecessor and successor relation between activities of the local effective event logs,and mining interface transitions based on activities which have a frequent relation of predecessor and successor with others,then consider all of the pre-set transition to output the initial transition of module nets and regard it as an input,add activities to it gradually to mining module nets effectively.(2)Proposing a decomposed mining method of process model based on behavior feature net to deal with event logs that contain a large number of different activities.Combining with the concept of the behavior footprint based on the behavior profile of the Petri net,mining the behavior matrix by analyzing the behavior relation between different activities,and computing the graph of behavior relation to divide clusters of activities.At the same time,discovering subnets by filtering sub-logs and mining behavior feature nets effectively based on each subnet.Finally,the complete net is formed by combining behavior feature nets of subnets.The method not only reduces the computational complexity of process mining effectively,but also improves the preciseness of decomposition mining algorithm that makes the mining algorithm better applicable to more fields.(3)When the model including information of time/frequency or model requirements which add the designated task to the model,existing methods of model optimizing are mostly based on the view of the configuration transition,optimizing the process model by mining the hide transition and block transition.However,the model requirements of model cannot be solved effectively through mining the configuration transition simply.Therefore,a model optimizing method based on segment adaptive is proposed in this paper to solve this problem.Firstly,the model requirements of the system operation feedback are analyzed based on configuration process model under the configuration optimizing and search the process segment that in need of adaptive.Secondly,the hide transition and block transition which are overlapping with adaptive segment will be replaced by this segment,and the others will be retained.Finally,selecting the appropriate adaptation pattern type based on the model requirements,and the configuration process model is optimized by adding the time task to the adaptation fragment or limiting the number of its occurrence by the adaptation pattern.