跨组织业务流程模型挖掘与质量评估

Discovery and Evaluation of Cross-organization Business Process Models

跨组织业务流程需要多个组织相互配合,协同工作来完成一项由单个组织无法完成的任务.由于跨组织业务流程的复杂性与分布性,其建模与分析过程是一项耗时且容易出错的任务,要求建模人员拥有丰富的经验和行业知识.流程挖掘通过分析业务信息系统执行过程中产生的日志为模型构建提供了一种自动化方法.然而,传统的流程挖掘技术仅支持单个组织的日志挖掘,无法有效地处理跨组织业务流程挖掘问题.本文针对此问题提出一种跨组织业务流程模型挖掘方法.首先扩展已有的流程挖掘方法来进行单个组织的业务流程模型挖掘;其次,定义组织间三种典型的协同模式,并提出相应算法以挖掘组织间的协同模式;再将各个组织的流程模型和协同模式集成,得到全局跨组织业务流程模型;最后采用传统的质量评估指标和提出的协同模式拟合度来量化发现的跨组织业务流程模型质量,通过四个不同的跨组织业务流程案例与已有挖掘方法进行比较,验证本文提出挖掘方法的有效性和可用性.

In a cross-organization business process,organizations need to collaborate with each other to complete missions that cannot be handled by a single organization.Due to the complexity and geographical dispersion of cross-organization business processes,the construction is a timeconsuming and error-prone task that requires practitioners to have extensive experience and business background.Process mining provides an automated method for model construction by analyzing event logs collected from the execution of business information systems.However,traditional process mining technologies only support model discovery of a single organization,and cannot effectively deal with the problem of cross-organization business process mining.To solve this problem,a cross-organization business process model discovery method is proposed.More specifically,it first extends existing process mining methods to discover business process model of a single organizations.Then,three typical collaboration patterns among organization and their corresponding discovery algorithms are introduced.Next,process models and collaboration patterns are integrated to obtain a global cross organization business process model.Finally,the traditional quality evaluation metrics and the fitness of collaboration pattern are invented to quantify the quality of discovered cross organizational business process models.By comparing with the state-of-the art process discovery techniques using four public cross organization business process cases,the effectiveness and applicability of the proposed approach is illustrated.