基于K-means和PLS-DA的期刊评价关键指标研究

目的/意义]期刊的评价指标和评价方法众多,研究期刊评价的指标重要性有重要意义。方法/过程]本文将期刊评价指标分为“事前”重要性与“事后”重要性,并重点研究“事后”重要性即客观聚类分析后的关键指标确定。以JCR2019年经济学期刊的文献计量指标为研究对象,运用K-means聚类得到评价结果分类,然后基于主成分分析、偏最小二乘法判别分析计算期刊评价的关键指标。结果/结论]使用多因素评价能够综合反映各变量的共同影响,使用“降维”思想能够较好保留原始变量的信息,降低多重共线性;PLS-DA方法得出影响期刊评价结果的最重要3个指标是期刊影响因子、他引影响因子和五年影响因子,并且在指标含义上是互补的;期刊影响因子、他引影响因子和五年影响因子的重要性相差不大,且期刊影响因子的重要性相对更大;PLS-DA对一般期刊判别效果很好,较好期刊判别效果一般,无法对A类期刊进行有效判别。

Key Indicators of Journal Evaluation Based on K-means and PLS-DA

Purpose/Significance] There are many evaluation indicators and methods for journal evaluation, and it is of great significance to study the importance of evaluation indicators of journals. This paper proposed an analysis framework for "post-event" analysis of the importance or weight of evaluation indicators. Method/Process] This paper divides the journal evaluation indicators into "before" importance and "after" importance, and focuses on the "after" importance, that is, the determination of key indicators after objective cluster analysis. Taking the bibliometric indicators of JCR 2019 economics journals as the research object, K-means clustering was first used to obtain the classification of evaluation results, and then the key indicators of journal evaluation were calculated based on principal component analysis and partial least squares discriminant analysis, and the importance of each indicator was analyzed. Starting from the objective results and the meaning of the indicators themselves, this study expounded the reasons why indicators are important. Results/Conclusions] Compared with the importance of "before", the importance of "after the fact" is to determine the evaluation results first, without involving weights, and completely based on data evaluation. There is relatively little controversy in the selection of methods. The indicator is more scientific and objective; the use of multi-factor evaluation can comprehensively reflect the common influence of each variable, and the use of the "dimension reduction" idea can better retain the information of the original variables and reduce the multicollinearity of the evaluation indicator; K-means cluster analysis methods, both the PLS-DA and PLS-DA models, are capable of evaluating and classifying journals. According to the results of PLS-DA, the VIP value of five indicators is greater than 1; the three most important indicators that affect the evaluation results of journals obtained by the PLS-DA method are the journal impact factor(IF), the other citation IF and the five-year IF, and the meaning of the indicators is the IF of other citations and the five-year IF make up for the shortcomings of the IF; the importance of the journal IF, the IF of other citations and the five-year IF is not much different, and the importance of the journal IF is relatively greater. In the quantitative evaluation of journals, it is necessary to use multi-index evaluation to make the evaluation more comprehensive and scientific; PLS-DA has a good effect on general journals, but the better journals have a general effect, and cannot effectively distinguish Class A journals.