Process Model Mining Methods For Special Loop Structures Via Petri Nets

With the development of computers and networks,it has become a new trend for enterprises to use information systems to replace traditional production management methods.These systems,while improving efficiency for the enterprise,also produced a large number of logs.The log is the product of recording how the system works and can provide some valuable information.This information is an important decision-making factor that drives process improvement.Therefore,mining this information from logs is one of the research hot-spots.Process mining is dedicated to building business process models from the logs generated by enterprise information systems.The process model is a formal reflection of business processes,which helps us to understand the business process intuitively.It is based on data mining and Business Process Management(BPM),and combines them to form a new discipline.Application scenarios for process mining include process discovery,mining include process discovery,conformance checking,and process enhancement.Among them,process discovery is the foundation of process mining research.There are still many challenges that need to be solved urgently.Mining process models that contain multi-parallel length-two loops of triangles and describing indirect dependencies in the process model are two of them.Based on Petri nets and association rules,two methods extended from(Alpha)? algorithm are proposed to solve the above problems.In order to mine the process model of multi-parallel length-two loops of triangles from incomplete logs,(AlphaMatch)?M algorithm based on Petri net is proposed.It is can mine parallel length-two loops of triangles in a log that does not contain continuous repeated sequences.The idea of classification-matching was used for this method.According to the relationship of the number of activities,the activities that constitute length-two loops are divided into two types,and then according to the relationship of the activity position,the two types of activities are matched using the location of the activity.This paper uses the footprint matrix to record the relationship between matched activities.Through the footprint matrix,a process model with multi-parallel length-two loops of triangles can be constructed.Aiming at mining the indirect dependency of activities,this paper extends the a algorithm and proposes the(AlphaTR)?TR algorithm to mine indirect dependencies in the model.This method was based on association rules and Petri net.First,the loop structure and choice structure were identified by ordering or quantitative relations.Then,the association rules of activities in the above structures were mined by building tuples.An extended Petri net model is proposed to describe the existence of indirect dependencies by combining association rules and Petri nets.Finally,a large number of experiments and multi-dimension analyses and comparisons were performed on the ProM platform to verify the correctness and effectiveness of our method.