一种基于隶属函数的相似度计算方法及其应用  

Similarity measurement method based on membership function and its application

作　　者：李军均[1] 戚进[1] 胡洁[1] 彭颖红[1] 

机构地区：[1]上海交通大学机械与动力工程学院,上海200240

出　　处：《计算机应用研究》2010年第3期891-893,903,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(50775140;50575142;60304015);国家"863"计划资助项目(2008AA04Z113);上海市科委基础研究重点项目(08JC1412000);上海市科委资助项目(09DZ1121400);国家教育部博士点基金资助项目(200802480036)

摘　　要：针对在模糊实例检索中传统的基于距离函数相似度算法的复杂性和不精确性等缺点,通过分析隶属函数提出了基于隶属函数的相似度计算方法。该方法根据不同类型的检索需求,构建了针对模糊数字型需求和模糊语义型需求的相似度计算模型;同时考虑了属性权重问题,采用基于相似度离差信息的客观赋权方法将主观权重和客观权重加以组合,以组合权重来计算实例的全局相似度。以汽车发动机的概念设计为例,验证了基于隶属函数的相似度计算方法能对不同类型的需求实现快速而有效的实例检索。It is very complex and imprecise that using distance functions to calculate similarity under fuzzy conditions. This paper introduced a novel similarity measurement method based on the analysis of membership functions. According to different types of demand in case retrieval, constructed respectively similarity measurement models for fuzzy numerical demand and fuzzy linguistic demand. In the weights assignment, this paper used the deviation information of similarity values to calculate objective weights, and then combined the objective weights and subjective weights to form synthesis weights. Finally, applied this method in the conceptual design of motor engines. It demonstrates that this method can perform rapid and effective case retrieval in response to different demand.

关 键 词：隶属函数 相似度 权重 实例检索 

分 类 号：TP391.72[自动化与计算机技术—计算机应用技术]