Hybrid clustering analysis for mapping large scientific domains

A hybrid clustering method combining cross-citation and textual analysis is applied to cluster more than 8,000 journals covered in the Web of Science ,2002-2006. Unlike in a previous study where we assumed 22 clusters for comparison with the 22 fields according to the classification scheme of Thomson's Essential Science Indicators, this study uses a 7 clusters solution, which is one of the candidate results obtained from the clustering process. Based on an agglomerative hierarchical clustering algorithm, all the journals under study have been clustered into 7 large scientific domains. The evaluation of the obtained clustering provides consistent results as considered from the cognitive perspective and the most characteristic terms, which are obtained from the textual component of the classification process, giving a clear description of each individual cluster. The cross-citation network visualises the citation relations among clusters and the asymmetric links reflect the direction of information flow among journals and clusters. Several indicators including PageRank, strong links and entropy are used to identify and analyse representative journals of each cluster.

Publication year:2009

BOF-keylabel:yes

IOF-keylabel:yes

Authors from:Higher Education