混合聚类方法用于科学结构研究--以“心理学、社会学和教育学”领域为例

Hybrid clustering for the analysis of subject-classification:a case study of "psychology,sociology & education" domain

在前期的相关研究中,一种基于"引文-文本"相似度的混合聚类方法被用来对2002-2006年期间Webof Science数据库中8305种期刊进行了聚类分析,得到22聚类和7聚类两种体系。与之前的宏观分析不同,7聚类体系中的一个特定的聚类,即"心理学、社会学和教育学"聚类,将被集中研究。该聚类的内部学科结构以及其在22聚类体系中如何分解为三个子聚类将被细致地讨论和分析。对聚类之间的互引网络以及聚类中不同类型的"重要期刊"进行了展示,此外,还进一步将该聚类中760种期刊的混合聚类结果与Thomson-Reuters公司的ESI学科分类进行了对比,基于比较分析的结果,对该领域内部的科学结构进行了量化解读。

In previous studies,a hybrid text/citation-based method has been used to cluster 8305 journals covered by the Web of Science database in the period 2002-2006.In particular,two classification schemes,respectively,22 and seven clusters system,have been studies in detail.Different from the previous macro-analyses,the present paper will focus on one selected cluster obtained from a seven-field solution,namely,the cluster of ＂Psychology,Sociology Education＂.The internal subject structure of this specific cluster and breakdown into three sub-clusters in the 22 clusters scheme are analysed.Furthermore,the comparison of the hybrid clustering of these 760 journals and Thomson-Reuters＇ ESI classification allows the validation and improvement of the latter ＂expert＂ classification scheme.