Botanical Garden, Ural Branch of the Russian Academy of Sciences, Ekaterinburg, 620144, Russia

org17@mail.ru
sash@botgard.uran.ru
mai@usfeu.ru


Abstract. The expediency of use of changes of attribute values at classification, which supposes account of various heredities, is proved. The classification meets difficulties, for object domains in a space of attributes are nonconvex, and overlap of domains is substantial. It is proposed to use functional dependences (including dynamic ones) between attributes to overcome the difficulties. Various ways of realization of such procedures, based on generalization of the concept of polythetic class, are considered.

Key words: classification, cluster analysis, regression analysis, result processing of morphometric measuring.