扩展ECA法中不可预测异常的相似实例检索研究  

Searching similar instance of unpredictable exception in extended ECA method

作　　者：陶亚雄[1] 王坚[1] 凌卫青[1] 

机构地区：[1]同济大学CIMS研究中心

出　　处：《计算机集成制造系统》2008年第6期1182-1188,1199,共8页Computer Integrated Manufacturing Systems

基　　金：国家科技支撑计划资助项目(2006BAF01A46);上海市基础研究重点资助项目(06JC14066)。~~

摘　　要：扩展事件-条件-活动规则法利用实例相似的概念,通过相邻算法在事件-条件-活动库中检索相似实例,来辅助处理不可预测的流程异常,但实例属性的加权系数难以准确确定,因而存在实施困难。为此,本文提出共轭相似度的概念及计算公式,借助相似度阈值和实例属性优先级,分两次检索相似实例,从而克服了加权系数的困难,简化了相似度计算、实例检索和判断过程,提高了系统处理不可预测异常的能力,提高了扩展事件-条件-活动法的有效性。Through concept of similarity between two instances and Nearest Neighbor （NN） algorithm, unexpected process exception could be handled referring to valuable means from its similar instance searched by extended Event- Condition-Action （ECA） rule method. However, it was hard to determine weighted factors of instance attributes which led to difficulty in implementation. Thus, concept of conjugated similarity degree and its calculating formula were defined. Recurring to threshold of conjugated similarity degree and priority of instance properties, the most similar instance in ECA rules base could be achieved after two times searching without any trouble in coefficients. This algorithm simplified similarity computation, case search and judgment process, and improved the system＇s ability to handle unexpected exception and the validity of extended ECA method was increased.

关 键 词：扩展ECA法 相似实例 相似度 共轭相似度 相似度阈值 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]