Abstract
Objective: In this paper, we propose a novel method to classify scientific journals into subject fields based on interjournal citation data. Object: Citation data of 5964 journals from JCR2004 Science Edition CD-ROM. Cited ratio vectors of the journals were used for analysis. Method: Our original clustering programs based on k-means were used. For the clustering, cosine distance (1-cosθ) was used. Results: In the cited-ratio-vector space, the spike-like-regions which appear to correspond to particular scientific fields were observed. And we succeeded to extract each spike-like-region by our clustering method. The number of fields (clusters identified) was 22. It was revealed that the cited feature of prominent multidisciplinary (general) science journals was similar to that of biochemistry journals.