| With the rapid development of modern business markets,business processes play an increasingly important role in the development of companies.They are often used for such purposes as supporting communication within the organization,documentation in projects and training of corporate employees.The wide range of application areas is accompanied by the existence of a large number of business process models.Typically,large enterprises often have process repositories consisting of hundreds or even thousands of models,and these large-scale process repositories are often developed by different people.The need for efficient application and management of these process repositories has become one of the indispensable technologies to enhance enterprise competitiveness,and such a need brings the challenge of accurate and efficient process model similarity metrics.Such techniques as process model similarity metrics have been used in many areas of business process management,for example,process model similarity metrics can be used to simplify the management and facilitate the reuse of process variants,merging process models,querying process fragments and process models,managing large-scale process repositories,automating process execution,consistency checking,and other scenarios.In order to further measure the differences between process variants and the base model based on distance,Chapter 3 of this paper proposes a distance similarity analysis method for process variants based on transition adjacency relations,which combines two similarity metrics,transition adjacency relations similarity and distance,and uses the ratio between them to measure the similarity value between the base model and the process variants.This similarity measure not only avoids the problem of one-sided similarity measure results when a single metric is used,but also solves the problem of distinguishing equidistant process variants in process variant clusters.In order to address the situation that most current process model behavioral similarity metrics algorithms may lead to a similarity metric result of 1 for two process models with differences in both behavior and structure when dealing with process models containing invisible tasks or cyclic structures.The fourth chapter of this paper takes the free choice workflow network as the basis for formal modeling,proposes the concept of task execution relationship,defines the direct following relationship,indirect following relationship and unreachable relationship between two tasks.Then we constructs the task execution relationship matrix based on the task execution relationship,and then calculates the process model similarity by measuring the similarity between two task execution relationship matrices.In Chapter 5 of this paper,a process similarity measure based on probabilistic BPMN process model is designed to address the similarity measure between BPMN process models,which assigns weights to the flow relations in the BPMN process model based on the probability of occurrence to obtain the probabilistic BPMN process model(Probabilistic BPMN Model).Based on the graph edit distance,the editing operation cost between different gateway nodes is further refined,and the similarity between two BPMN process models is measured by the graph edit distance of the PBMN process model,which makes up for the current deficiency of insensitivity to changes in the gateway of the process model when using the graph edit distance to measure the similarity value of two BPMN process models.In Chapter 6 of this paper,two process similarity measures are designed,which are divided into two scenarios:(1)Improving the business process diagram when both the event log and the process model are known,proposing the concept of improved business process diagram,and using the frequency information in the event log to assign weights to the improved business process diagram to obtain a weighted improved business process diagram,and then based on the weighted improved business process diagram editing operation to measure the similarity results;(2)in the case where only the event log is known,the event log is abstracted into a probability distribution by combining the activity information in the event log with the execution frequency,and then the similarity between two probability distributions is measured based on the JS divergenceThere are four experiments in this paper,which are used to verify the process variant distance similarity analysis algorithm based on variant immediate neighbor relationship,a process similarity measure using task execution relationship,a process similarity measure based on probabilistic BPMN process model,and a process similarity measure based on event log and model structure.The results of the four experiments mainly validate the feasibility of the four main algorithms proposed in this paper.In addition,the superior performance of the proposed algorithms in terms of accuracy is also verified by two benchmark data sets.In summary,the main research contents and contributions of this paper are as follows.(1)A distance similarity analysis algorithm for process variants based on transition adjacency relations is designed for handling the problem of differentiating between equidistant process variants.(2)A process similarity metric using task execution relationships is designed,which can effectively handle invisible tasks as well as cyclic structures with high query accuracy.(3)A process similarity measure based on probabilistic BPMN process models is designed for the similarity calculation among BPMN process models.(4)A process similarity measure based on the event log and model structure of the process model is designed,and the algorithm proposes two different process model similarity measures based on whether the process model is known or not when the event log is known.Figure [26] Table [26] Reference [118]... |
PDF Full Text Request