摘要
通过对主成分分析法(PCA)基本算法的分析,从企业绩效综合评价的角度出发,指出了在指标变量的多重相关性处理、定性指标定量化、原始数据无量纲化等方面存在的弊端,引入了分级指标体系的设计方法、数量化理论(Ⅲ)、分类无量纲化方法解决之。针对PCA中特征向量方向的确定问题,提出了特征向量方向确定的原则。针对主成分的提取和权重赋值问题,提出了优先递阶提取法和两两比较判断矩阵法解决之。最后,针对线性PCA的结构化问题,探讨了当今研究的方向和思路。
On analyzing the basic arithmetic of the Principal Component Analysis (PCA), this paper refers to some flaws exist with such aspects as handling the multiple correlation of variables, qualification of quantitative indicators, and dimensionlessness of original data and then comes up with the solution with the help of classified index system, quantification theory (III), and dimensionless classification. The direction of characteristic vector in PCA is also brought under some guidance and the article is drawn to a close with some ideas about the research direction on the picking up of principal components and their power evaluation.
出处
《山西财经大学学报》
北大核心
2004年第4期80-85,共6页
Journal of Shanxi University of Finance and Economics
基金
国家自然科学基金资助项目(70272029)
国家社会科学基金资助项目(01BJL022)
教育部人文社会科学十五规划第一批项目(01JD790021)。
关键词
绩效评价
主成分分析
优先递阶提取法
performance evaluation
principal analysis
prior sequence extraction
