科学计量学常用工具的聚类算法分析

Analysis of Clustering Algorithm in Software of Scientometrics

摘要:

在进行科学计量学研究的过程中,聚类分析是进行科学计量研究的常用方法。面对纷繁复杂的聚类算法,确定聚类方法的划分质量以获取最佳聚类数目是进行聚类评价的重要研究内容。为此,我们从科学计量学研究的共现矩阵数据出发对常见的聚类分析工具算法进行了分析。首先,介绍了聚类算法的过程以及聚类有效性指标的分类；然后,分别评述了科学计量学常用软件中的几种聚类算法；接着,分析聚类算法的特性并采用基于类内紧密度和类间分离度对聚类结果的有效性进行探讨；最后,总结各聚类算法的效果并对应软件分析的结果进行案例分析。

Abstract:

During the course of scientometrics study, cluster analysis is a common method for scientific research. In the face of complex clustering algorithms, the quality and the optimal number of the clustering method is very important for clustering evaluation. Therefore, based on the co-occurrence matrix data of scientific research, the common clustering algorithms and tools are analyzed. First, this paper introduces the classification of clustering algorithm and clustering validity index; then, several clustering algorithms of measurement software commonly used in scientometrics are reviewed; then, analysis the characteristics of clustering algorithm, and the validity of clustering results is discussed based on the compactness and separation of the clusters. Finally, the effect of clustering algorithm and the result of software analysis are summarized.