Hierarchical visual data mining for large-scale data (Q1887250)

The authors discuss some new techniques of visualization for large data sets (with millions of cases and hundreds of variables). They are mainly based on some previous clusterization algorithms. Scatterplots and parallel coordinate plots for clusters are described where the clusters are visualized by the mean-band method. A structure-based brushing is considered by which the user can explore a hierarchical clusterization tree. It is proposed to use ``dimension clusters'' when the data are multidimensional. The public domain visualization package XmdvTool implements all theses techniques.