一种等价类区分的多属性评价指标约简算法  

Eliminating the Correlation of Multi-Attribute Evaluation Based on Equivalence Class Partition

作　　者：朱海萍[1] 王云岚[2] 李增智[1] 

机构地区：[1]西安交通大学计算机系统结构与网络研究所,西安710049 [2]西北工业大学高性能计算研究与发展中心,西安710072

出　　处：《西安交通大学学报》2009年第4期26-29,共4页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(90304006,60673170)

摘　　要：针对多属性评价中消除评价指标之间相关性的问题,从指标的组成出发,将评价指标分为元指标和合成指标,并分别给出了线性合成指标和非线性合成指标的相关度计算方法.借鉴Dijkstra算法,提出了基于指标等价类的合成指标动态约简算法,即通过等价类划分来缩小参与约简的指标范围,并充分利用前一次约简以求得最终指标约简的结果.仿真实验表明,随着指标数目的增多,所提算法比直接使用改进Dijkstra算法进行指标约简的运行时间要少,可提高大规模指标约简的执行效率.To eliminate the correlation among evaluation indexes, the evaluation indexes are divided into element indexes and synthetic indexes. Then the methods of calculating the correlation degree of linear synthetic indexes and nonlinear synthetic indexes are given. An algorithm of eliminating the correlation of synthetic indexes based on index equivalence class is proposed in terms of Dijkstra algorithm. The reduction ranges of evaluation indexes can be reduced, and the anterior reduction results are used to increase the efficiency. Comparisons with the modified Dijkstra index reduction algorithm show that the proposed algorithm based on equivalence class spends less running time. The more indexes there exist, the more efficient the proposed algorithm is.

关 键 词：多属性评价 约简算法 相关度 指标等价类 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]