摘要
将科技评价中几种常见的效用函数合成方法进行了比较,重点分析了不同合成方法的理论意义。认为加法合成比较适合不同类并且可以互相替代的指标的合成;对数合成比较适合有相乘关系或先后次序的指标的合成;调和平均比较适合同类指标的合成,往往用在二级或三级指标的合成当中。在同一评价中,由于一级指标之间、二级指标之间存在的关系往往不同,因此应该灵活选择不同的合成方法进行处理。平方平均合成与乘法合成要谨慎选取。本文还修正了传统加权调和平均合成公式的一个错误。在此基础上,以汤姆森路透(Thomson Reuters)科技集团JCR期刊中的电气期刊为例进行了分析。
This paper compares some common utility function compound methods in science and technology evaluation and analyzes their principles.The results show plus compound method is suitable for the same and replaceable indicators,logarithmic compound method is suitable for ranking and multiplying indicators,and weighted harmonic average method is suitable for congeneric indicators.Different compound methods should be selected flexibly in the same evaluation because of the different relationships between the first and second level indexes.Square average method and multiply compound method should be selected more carefully.The paper also corrects an error of weighted harmonic average and gives an example of the integrated application of different compound methods based on JCR of Thomson Reuters.
出处
《情报杂志》
CSSCI
北大核心
2012年第11期77-82,70,共7页
Journal of Intelligence
基金
教育部人文社会科学青年项目"科技期刊评价中利益相关者的关系特征分析及其改进"(编号:12YJC870024)
关键词
科技评价
效用函数
合成方法
science and technology evaluation utility function compound method
