摘要
案例推理(CBR)系统可用于支持半结构化的决策问题,它的有效性取决于其案例检索能力,而案例检索性能的关键是案例匹配。目前的符号型匹配如最近相邻匹配假设描述符是相互独立的,其重要性是预先定义的,这和半结构化问题中描述符之间会有复杂关系且权重无法明确定义的特点不符。本文就从介绍CBR技术、案例检索出发,将NN匹配函数和一种分布式表达的关系型CBR结构进行比较,希望有助于提高案例检索的性能。
The goal of case- basedreasoning system is to support ill - struc-tured decision problem. Its effectivenessdepends on its retrieval performance, andthe matching process is essential to the re-trieval performance. Existing symbolicmatching functions such as nearer -neighbor (NN) matching function assumethat the descriptors of a case are indepen-dent and their weights are pre - defined,which is inconsistent with the fact in ill-structured decision problem that descrip-tors may have complex relationships andit's difficult to define their importance ex-plicitly. This paper first introduces CBR,retrieval and matching process, then a dis-tributed connectionist CBR architecture,which can improve the retrieval perfor-mance in ill-structured decision environ-ment.
出处
《管理信息系统》
1999年第5期37-41,共5页
Management Information Systems China
关键词
人工智能
案例推理系统
案例检索
case -based reasoning
case retrieval
matching
ill-structured decision problem
descriptor
