The technology of hierarchical agglomerative cluster analysis in library research

Abstract

This article describes the simple technology of the hierarchical agglomerative cluster analysis of 20 different libraries, presented by the samples of classification attributes of the same volumes. It is necessary to construct a proximity matrix for an effective process of cluster analysis, using the data from the table “library-classification features”. For separating a set of selected objects into clusters, so that each of them has objects, the most appropriate for its type, it is necessary to create a table “object-property”, where libraries are objects and individual and equal dimension vectors (sets) of characteristic classification features are properties. To do this you should: form the set of libraries that are the objects of clustering; define for each library the set of classification features and its power (volume) in the same nominal scale; choose the value scale of classification features; form a table object-property. This technology is implemented in the environment of MsExcel-2003. It includes the transformation of one-dimensional data into multi-dimensional indexes, using the descriptive statistics and distributions of individual parameters, the creation of the “object-property” table, the building of the proximity matrix, the definition of the dendrogram structure and the cluster interpretation. The application of clustering method can further focus on creating algorithms of effective information search, and also building the scientifically-reasonable classification systems orientated on the library science. The method of hierarchical agglomerative cluster analysis can be used with typological or semantic distribution of library funds, or studying of their thematic and specific composition. This method can be considered as universal, that gives an opportunity to formalize the typology division of any objects of librarianship.