流程增量挖掘中的模型更新方法  

Incremental Process Mining with New Model Update Method

作　　者：马慧[1] 汤庸[1] 吴凌坤[1] 

机构地区：[1]中山大学计算机科学系,广州510275

出　　处：《计算机科学》2009年第5期154-157,共4页Computer Science

基　　金：国家自然科学基金(60673135;60373081)重点项目(60736020);教育部新世纪优秀人才支持计划(NCET-04-0805);广东省自然科学基金(7003721)资助

摘　　要：正确发现流程实际运作情况对工作流管理有着重要的意义。流程挖掘抽取系统日志信息,挖掘流程的真实运作模型。目前很多该方面的研究,着重于从一份日志中挖掘出工作流模型。然而,这些挖掘方法只关注日志信息,忽略了流程设计者的先验知识。而且,日志所包含信息量较大,进行一次挖掘耗费较大。因此,希望能结合已有工作流模型及新增日志信息,更新工作流模型。已有研究给出对模型及日志的增量挖掘算法。但是,业务流程会随着时间推移变更,可能已有的任务被取消了,因此在新增的一段日志中该任务没被记录。但由于该任务曾经在已有日志中记录下来,故应用已有挖掘算法或增量挖掘算法,在更新模型中,该任务也会被挖掘出来。提出了一种增量挖掘模型更新的改进算法。通过流程设计者的先验知识及统计任务出现的频率,判断该任务是否被取消。最后给出一个实验,验证算法的可行性。A thorough understanding of the way in which a workflow process is executing is essential to workflow management. By extracting information from work traces, such as system log data, process mining aims to discover the actual behavior of a workflow process. Current researches mainly stress on discovering a workflow model from a whole log. These process mining approaches only use the log＇s information, while losing sight of the process disigners＇ prior knowledge. Besides, the large volumn of data which the log contains makes the process mining a time-consuming job. It is expected that the new model is derived by conbining information of the existing model and the new incremental log. The problem of incremental process mining was studied. However,workflow process may change as the time goes by. It is possible that a certain task has been canceled. Though it is not recorded in the incremental new log, it has been recorded down in the old ones. Thus by applying current process mining algorithms, or the proposed incremental process mining method, the new workflow model still contains the cancelled task. An improved incremental model update method was proposed. Whether a task is cancled or not is decided by prioror knowledge and its execution frequency. An experiment was given to show the validity of this method.

关 键 词：流程挖掘 增量挖掘 工作流模型 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]