Research On Several Key Problems Of Artifact-centric Business Process Model Mining

With the advancement of Business Process Management technologies, organizationsand enterprises are accumulating large repositories of business process models, whichbecome precious data assets that supports process integration, process reengineering,process optimization and other business analysis and process design activities. Hence, aneffective method to manage and utilize this knowledge is becoming a critical issue andprocess model mining technology arises at this moment to tackling this problem.Currently, studies on business process management are experiencing a transition fromcontrol-centric to data-centric approach which is notably marked by the Artifact-centricapproach. Therefore, Artifact-centric business process model mining is becoming achallenging task.This thesis implements researches on some key issues of process model miningtechnology such as process model similarity measuring, process model clustering, servicecomposition mining and Artifact behavioral conformance checking. The completeresearch is outlined below:Firstly, a bipartite graph for artifact-centric business process models and its graphmatching algorithm are presented. As the ordinary graph can not completely describe therelevance between Artifact and its processing services, a bipartite graph forartifact-centric business process models is proposed, which transforms process modelmatching to graph matching. Meanwhile, a similarity matching algorithm namedArtiMatch is given.Secondly, a novel Artifact-centric business process behavior similarity measuringmethod is proposed. The currently existing behavior similarity measuring methods cannot be well applied to artifact-centric business process due to the Atifact lifecycle features.Therefore, an Artifact lifecycle tree model is proposed and the process behavior modelsimilarity is calculated by measuring service dependency relations and the similarity ofattribute assignment sequence in the service executing path.Thirdly, an Artifact-centric business process model clustering method is presented. Usually, the quality of a clustering result is determined by the extraction of characteristicvalues of process model similarity. On account of this, this thesis proposed a processmodel clustering and a model matching algorithm based on the similarity value of keyArtifact, process structure and process behavior by using hierarchical clustering technics.Fourthly, a service composition mining approach for artifact-centric businessprocess models are proposed. Traditional approach on service composition miningoverlooks the relationship between data and its services. According to the correlationbetween services and Artifacts, a new approach for mining service composition patternsbased on Artifact-aware aspect is also presented using apriori paradigm.Finally, a comprehensive approach for checking behavior conformance of Artifact isproposed. Existing conformance checking techniques has encountered some deficienciesfor that Artifact is composed of data schema and its lifecycle schema. Therefore, a novelapproach for checking behavior conformance of Artifact according to its lifecycle ispresented in a perspective of Artifact lifecycle pattern. In this approach, the problem ofbehavior conformance checking is transformed into the decidability problems oflanguage. Meanwhile, a Turing machine model is designed for validating the language.Lastly, the calculation method of fitness metric in conformance is also presented.