【作者】 高昂；

【导师】 杨扬；

【作者基本信息】 北京科技大学 ， 计算机应用技术， 2010， 博士

【摘要】 工作流管理系统是利用计算机技术来自动化地执行组织机构中业务流程部分或全部的一种技术，能大大提高业务流程的处理效率和执行质量。由于流程的不确定性，大大增加了工作流建模的难度。而工作流挖掘技术则颠覆了工作流模型生命周期过程，收集工作流模型运行阶段的信息并支持模型的再设计，因此工作流挖掘在工作流领域中显得日益重要。在分析工作流挖掘技术起源和发展的基础上，对其研究现状进行了总结，指出了现有的工作流挖掘技术的一些不足。针对这些不足在工作流控制流视角，案例视角，组织结构视角的挖掘方面分别提出了新的挖掘算法与解决方案。主要研究成果如下：1)在工作流控制流视角的挖掘方面，提出了一种新的工作流控制流模型挖掘算法。首先给出了工作流模型直接依赖矩阵的定义，并利用工作流日志建立直接依赖矩阵的算法，设计了一种基于直接依赖矩阵的模型自动生成算法，通过实例说明这种算法在提高模型准确率与描述能力等方面的优势。此外，为了提高工作流对动态不确定的适应能力，引入了工作流频繁模式。对频繁模式给出了定义，并在Apriori算法基础上提出了一种工作流频繁模式挖掘算法。2)在工作流案例视角的挖掘方面，分析了工作流日志中的数据属性如何影响工作流实例的路径选择。基于算法挖掘工作流日志过程模型，对过程模型中的决策点进行分析，通过决策树分析技术结合工作流日志中的数据属性挖掘出影响工作流实例路由的决策规则。分析了现实应用中决策规则挖掘所遇到的问题，并提出解决算法。最后通过测试程序测试并验证了挖掘过程。测试结果表明该算法能够正确地挖掘出决策规则。3)在工作流组织结构视角的挖掘方面，分析了工作流日志中的活动执行者的信息，并引入多种距离公制用以刻画日志中活动执行者间的各种关系度的强弱，提出了组织结构的四种挖掘方法：“默认挖掘”、“基于因果关系公制的挖掘”、“基于活动相似度的挖掘”和“基于过程实例相似度的挖掘”。通过以上方法对工作流日志进行组织结构挖掘得到组织网络，组织网络能够客观地反映出参与工作流程的组织实体并正确表示出组织结构。通过一个挖掘实例对挖掘过程进行了验证和说明。4)设计并初步实现了一个用于对工作流日志进行挖掘与分析的原型系统LogMiner。对LogMiner的设计架构进行了详细说明。对一个大型国有企业的工作流日志进行了详细的挖掘与分析。从挖掘结果中发现了企业业务处理流程中存在的问题，找到了制约业务发展的瓶颈，并提出相应的建议与改进方案，最终提高了企业业务流程的处理能力,提升了工作效率。更多还原

【Abstract】 Workflow is a technology imposing computer technique to execute parts ofor the whole business process within organization automatically. Through thedisposing efficiency and executing quality of process can be improved greatly.However, the complexity of the workflow management system increases thedifficulty of the workflow modeling. The goal of workflow mining is to reversethe process and collect data at runtime to support workflow design and analysis.So workflow mining technology which can improve efficiency and rationality inworkflow management system become more and more important in the workflowmethodology.Based on current research on workflow mining technology, origin ofworkflow mining technology and related evolution, some shortcomings needed tobe overcome in this field is pointed out. To provide solution for some of theproblems that described above, new algorithms and solutions for miningworkflow models are proposed in the workflow control-flow perspective、caseperspective and organizational perspective respectively. The main research resultsare as follows:1) In the workflow control-flow perspective field, a new algorithm formining workflow models is proposed. First, the definition of Workflow-modelDirect-depend Matrix(WDM) is proposed, and an algorithm is designed to set upWDM from workflow logs, at the same time, another algorithm is designed toautomate the modeling based on WDM. And according to the given example, thesignificant advantage of this algorithm will be found in improving accuracy,enhancing description ability. In another side of control-flow mining technology,to improve workflow for adapting to the uncertainty, the workflow frequentpattern is used. First, the frequent pattern discovery problem is defined, and afrequent pattern mining algorithm based on Apriori algorithm is proposed.2) In the workflow case perspective field, analyzed how data attributesinfluence the choices made in the process instances based on workflow logs.Mining process model in workflow logs based on-algorithm,analyzed thedecision point in process model, detection decision rule that affected the routingof a case through techniques based on decision tree analyze and data attributes inthe workflow logs. Discussed problems related to real-life decision rule mining,an improved algorithm is presented. To verify the presented ideas a test program has been implemented. The test result shows the alogrithm can correctly miningthe decision rule from workflow logs.3) In the workflow organizational perspective field, define metrics toestablish the relationships among originators by analyzing the information in theworkflow logs. Four methods for mining organizational structure from workflowlogs are presented in this paper, default mining、mining based on the causality、mining based on the similarity of activities and mining based on the similarity ofcases. By applying these methods, can construct the corresponding organizationalnetwork which reflects the organizational entities involved in the workflowprocess and corrects presented the organizational structure, and give an exampleto explain it.4) Based on the above research work, a workflow logs mining and analysissystem LogMiner has been designed and realized. The design architecture of theLogMiner is described in detail. For a large enterprise workflow logs carried out adetailed mining and analysis. The problems of business processes and bottleneckrestricting the business development are found from the mining results, and makeappropriate recommendations to improve the business processes, the enterprisebusiness processes processing capacity and work efficiency have been improved.更多还原