Research on Scientific Bibliography Clustering Technology Based on Citation Information Merging

Clustering analysis is an important step in extracting and analyzing technology bibliography trend. The bibliographic records, consisting of the title, authors, keywords, and publications, are short texts carrying less information than long texts, while bibliographic network has the citation relations among bibliographies. The experimental dataset therein refers to bibliographies in the computer field crawled from the existing literature databases. Based on traditional Hierarchical Clustering algorithm, this paper merges the cocitation relations. The experiment result shows that the mean silhouette coefficient of the Agglomerative Clustering algorithm that merges the co-citation relations among the bibliographies is improved obviously, thus effectively improving clustering result among bibliographies with strong co-citation relations, and good clusters provide a solid research basis for the follow-up bibliography trend analysis.