Classification of Hyperlink Networks by Multivariate Analysis of Network Indicators: The Case of Japanese Public Libraries
Classification of Hyperlink Networks by Multivariate Analysis of Network Indicators: The Case of Japanese Public Libraries
复制标题
通过网络指标多元分析对超链接网络进行分类：以日本公共图书馆为例
DOI：
--
复制
发表时间：
2009
期刊：
影响因子：
--
通讯作者：
F. Yoshikane
中科院分区：
文献类型：
--
作者：
Shuntaro Kawamura;Takafumi J. Suzuki;F. Yoshikane
本研究通过对网络指标进行多变量分析，对日本公共图书馆网页构成的47个县内超链接网络进行了分类，并阐明了网络的特征。该方法为公共图书馆网络分类和服务评价提供了一种新的方法。我们选择了九个网络指标，代表网络的五个方面：子组，每个节点的位置，距离，聚类和度分布。我们对这些指标进行了两种多元分析：主成分分析和聚类分析。结果表明，节点位置和度分布是影响分类效果最强的两个方面。都道府县内的网络分为三种类型：（1）连接较弱的网络;（2）有多个图书馆但只有一个中央图书馆的网络;（3）连接较多但集中在多个图书馆的网络。这表明，在一些县内网络中，市图书馆和县图书馆发挥着核心作用。
In this study, we classified 47 in-prefecture hyperlink networks formed by the Web pages of Japanese public libraries by applying multivariate analysis to network indicators, and clarified the characteristics of the networks. Our method provides a new way of classifying networks and examining the services of public libraries. We selected nine network indicators representing five aspects of networks: subgroups, position of each node, distance, clustering and degree distribution. We carried out two multivariate analyses of these indicators: principal component analysis and cluster analysis. The results indicate that the position of each node and degree distribution are the aspects that affect the classification most strongly. The in-prefecture networks can be classified into three types: (1) those that have only weak connections; (2) those that have many libraries and only one central library; and (3) those that have many links which are centralized in not one but some libraries. This indicates that in some in-prefecture networks municipal libraries as well as prefectural libraries play a central role.