Listy Biometryczne - Biometrical Letters Vol. 38(2001), No. 2, 67-79
|
AN INTERESTING GROUPING OF POLISH VOIEVODESHIPS OBTAINED BY SELF-ORGANIZING MAPS Anna Bartkowiak Institute of Computer Science, Wrocław University, Przesmyckiego 20, Wrocław 53-502, Poland |
We consider a data set containing 49 Polish administrative units (voievodeships), each characterized by 9 variables. To discover similarities and dissimilarities among the units we use self-organizing maps. Applying this method we obtain a very interesting map reflecting both geography, history and development of the units: we obtain grouping due to cultural development (university cities) and due to history, where the influence of powerful neighbours occupying Polish territory for over 150 years in the past is still visible. The method is compared with other methods of visualizing multivariate data: the grand tour and spin plots in principal coordinates.
visualization of multivariate data, self-organizing map, clustering of objects
