Research On Semantic Workflow Reuse Based On Case-Based Reasoning And Behavioral Characteristics

The efficiency of the business process operations is one of the key factors for modern enterprises and organizations to maintain the competitive advantages.With the wide application of business process management(BPM)technology,a large number of business process models have been accumulated in every industry or trade.These business process models are becoming important data resources.How to make full use of these data resources to complete the tasks of modelling new business processes is a challenge that the business process managers of modern enterprises and organizations need to face.The task of modelling new business processes can be achieved by reusing existing similar business processes or parts of them.Semantic workflows are workflows based on domain knowledge,which contain control-flow,dataflow and other constraints.They are being widely used in E-commerce,medical,software development,scientific analysis,information integration and even cooking.They provide good fundamentals for workflow reuse,and also bring challenges.According to the limitations of existing semantic workflow reuse,this dissertation proposes a method of semantic workflow reuse based on behavioral characteristics.On the basis of above studies,a theoretical framework of semantic workflow reuse is established based on case-based reasoning and the behavioral characteristics of semantic workflows,in order to improve the efficiency of semantic workflow reuse and further promote the development of workflow reuse technology.The main content of this dissertation is summarized in four parts as follows.1.Two behavioral similarity algorithms for semantic workflows are studied.Existing similarity measures for semantic workflows only focus on structural characteristics of workflows,thus the similarity between semantic workflows with similar execution behaviors and different structures cannot be accurately assessed.A behavioral similarity algorithm is proposed based on the representative traces(RT).A type of behavioral characteristics,the set of RTs,is obtained to approximately represent the execution behaviors of the semantic workflows,and the behavioral similarities among semantic workflows are calculated with the similarities among the sets of RTs.Compared with the existing similarity measures for semantic workflows,the behavioral similarity algorithm based on RT achieves better similarity results.For the difficulty of reflecting the task occurrence relation between task nodes with RTs,a behavioral similarity algorithm is proposed based on the semantic task adjacency relation(STAR)with importance.A type of behavioral characteristics,the set of STARs with importances(ISTAR)is obtained to approximately represent the execution behavior of the semantic workflow,and the behavioral similarities among semantic workflows are calculated with the similarities among the sets of ISTARs.Compared with the behavioral similarity measure based on RT,the method proposed obtains more accurate similarity results.2.A retrieval algorithm for similar semantic workflows is studied based on behavioral characteristics.Existing retrieval algorithms for similar semantic workflows only focus on the structural characteristics of workflows,so only semantic workflows with similar structures to the query workflow can be retrieved.Thus the overall quality of retrieved similar semantic workflows is reduced.A two-phase retrieval algorithm for similar semantic workflows is presented based on the behavioral and structural characteristics of semantic workflows.The index tree of task adjacency relation(TARTree)and the index of data are constructed based on the domain ontology.For the query semantic workflow,first,the TARTree index and data index are used to filter the semantic workflow repository for semantic workflows candidates in order to reduce the computations of verification with the graph mapping similarity.Then,these candidates are verified with the graph mapping similarity,and the set of the ranked retrieved workflows is obtained as a result.Compared with existing retrieval algorithms,the method proposed improves the retrieval performance of similar semantic workflow at an acceptable cost,and can provide semantic workflow candidates with higher quality for semantic workflow reuse.Aimed at the requirements of retrieving semantic workflows which meet the given execution trace or partial execution trace,a two-phase retrieval algorithm for meeting semantic workflow based on the trace pattern is put forward.The trace pattern is defined to represent the given execution trace or partial execution trace,and the task node index is established.For the given query trace pattern,first,the set of semantic workflows candidates containing the identical or similar task nodes within the trace pattern is filtered from the workflow repository with the task node index.Then,the candidates are verified to ensure the each candidate meets the execution order of the mapped tasks nodes in the trace pattern based on task reachability relation.The experimental evaluation shows that the algorithm proposed can satisfy the retrieval requirement of semantic workflows meeting the given execution trace.3.The adaptation algorithm for retrieved semantic workflows based on the behavioral characteristics of semantic workflows is studied.Existing adaptation algorithms for retrieved semantic workflows require the predefined adaptation knowledge that is hard to be obtained,or are realized based on the structural characteristics of the reusable segments of semantic workflows(called workflow streams),but the quality of adaptated semantic workflows is not high.To solve the problem,an adaptation algorithm is proposed based on the behavioral characteristics of workflow streams.The semantic workflow repository is decomposed into a workflow stream repository.The data index of the anchor set is constructed and the matching rules of streams are designed.The behavioral similarity measure for workflow streams is also presented.First,the retrieved semantic workflow is decomposed into the set of workflow streams.For each stream in the retrieved workflow,the data index,matching rules,and the similarity measure are used to retrieve the matching stream that is most simlar to it,while the matching stream is ensured to have the highest consistency with the change request.Then,each matching stream is used to substitute the corresponding stream in the retrieved workflow,the adapted workflow is obtained as a result.Compared with the adaptation algorithm based on the structural characteristics of workflow streams,the algorithm proposed improves the average quality of the adapted semantic workflows,and improves the applicability and robustness of the adaptation algorithm,and provides a reliable guarantee for the efficiency of the semantic workflow reuse.4.The method of maintaining the semantic workflow repository with three types of indexes is studied.The increasing size of workflow repository will reduce the retrieval quality and efficiency of similar semantic workflows,which reduces the quality of adapted semantic workflow and increases the cost of semantic workflow reuse.Considering the characteristics of semantic workflow repository,a method for the maintenance of semantic workflow repository is proposed based on three types of indexes.Two types of indexes,the index based on behavioral characteristics and the output data index are constructed,and the index based on structural characteristics is introduced.The three types of indexes focus on the semantic workflow repository from different perspectives,and many query requirements can be met with the combination of these three indexes.The experimental evaluation shows that the method proposed provides an effective management for the increasing semantic workflow repository and a necessary fundamental to guarantee the efficiency of semantic workflow reuse.