主成分与因子分析在期刊评价中的改进研究

The Misunderstandings and Optimization of Principal Component and Factor Analysis in Journal Evaluation

分析了主成分分析和因子分析在期刊评价中的误区,并提出了优化方法。研究表明,无论是主成分分析还是因子分析,其隐含的假设是评价指标必须服从正态分布,在期刊评价指标普遍呈幂律分布的情况下,采用主成分分析和因子分析要慎重,应将评价指标取对数后再进行评价。采用主成分和因子分析即使评价方法相同,不同评价也不具有可比性。主成分或因子分析采用方差贡献率作为权重值得商榷,应结合专家打分来赋予权重。

This paper analyzed the misunderstanding of principal component analysis and factor analysis in the journal evaluation, and pro-posed optimizing methods. Research showed that, for both the principal component analysis and factor analysis, the implicit assumption is that evaluation index must obey normal distribution; in the situation that periodical evaluation index had a general“power-law” distribu-tion, using principal component analysis and factor analysis must be careful, and the evaluation index should take the logarithm after evalu-ation. When using principal component analysis and factor analysis, even if with the same evaluation method, different evaluation is not comparable. And for giving weightiness, it is questionable for principal component or factor analysis to use the variance contribution rate, it should be combined with expert scoring.