Classification and visualization of the social science network by the minimum span clustering method

Abstract

We propose a minimum span clustering (MSC) method for clustering and visualizing complex networks using the interrelationship of network components. To demonstrate this method, it is applied to classify the social science network in terms of aggregated journal-journal citation relations of the Institute of Scientific Information (ISI) Journal Citation Reports. This method of network classification is shown to be efficient, with a processing time that is linear to network size. The classification results provide an in-depth view of the network structure at various scales of resolution. For the social science network, there are 4 resolution scales, including 294 batches of journals at the highest scale, 65 categories of journals at the second, 15 research groups at the third scale, and 3 knowledge domains at the lowest resolution. By comparing the relatedness of journals within clusters, we show that our clustering method gives a better classification of social science journals than ISI's heuristic approach and hierarchical clustering. In combination with the minimum spanning tree approach and multi-dimensional scaling, MSC is also used to investigate the general structure of the network and construct a map of the social science network for visualization.

The study of subject-classification based on journal coupling and expert subject-classification system

“…In previous studies, clustering methods used in subject-classification analysis can be divided into three categories, methods grounded on multivariate statistics, such as principal component analysis (or factor analysis) (Leydesdorff and Cozzen 1993;White and McCain 1998;Leydesdorff 2006). Methods using the classical clustering analysis, for instance, hierarchical clustering (Zhang et al 2010(Zhang et al , 2012Braam et al 1991;Kronegger et al 2013;Ahlgren and Colliander 2009), minimum spanning tree (Chang and Chen 2011), etc. The last one are methods belonging to the clustering of social network in graph theory (Chen et al 2010;Qiu and Liu 2014;Waltman and Van Eck 2012;Leydesdorff and Rafols 2012;Börner et al2012;Gómez-Núñez et al 2014).Clustering methods based on multivariate statistical theory take the node pair which has citation behavior as the variable and case respectively, then clustering the cases with same characteristics using the idea of projection, but there is no definite standard on the division of clusters, this is also true when it comes to the choosing of cluster number, and the results of such methods can hardly form a clear hierarchical subject-classification system.…”

Visualization and analysis of SCImago Journal & Country Rank structure via journal clustering

“…They executed various clustering algorithms to detect groups of similar journals and put them into different well-defined clusters. Chang and Chen (2011) (Blondel et al, 2008) executed on journal cross-citation and hybrid matrices (Janssens et al, 2009;Zhang et al, 2009). In one of these publications (Zhang et al, 2010) the same methodology was applied at the ISI category level, i.e.…”

Visualizing the world's scientific publications

“…Graph‐based data mining has been useful in analyzing and comparing scientific domains (Quirin, Cordón, Vargas‐Quesada, & de Moya‐Anegón, ). Chang and Chen () have used the MSC method for clustering the social science network, providing an in‐depth view of the network structure at various characteristic resolutions. The clustering results of scientific journals from the aforementioned studies are helpful in understanding mutual interactions between various knowledge domains.…”