Q uniwersalna miara odległości
Q universal distance measure

Summary/Abstract: In clustering tasks we usually have to use variables which are measured in different scales. One of possible strategies is to cluster objects separately for each scale and then try to combine them into final partition. The second strategy is to use the distance measure which allows for calculation of distances in the classification space formed by attributes measured in different scales. This second approach is possible while applying Walesiak GDM distance. In the paper we go beyond the classical four scales – nominal, order, interval and rational. Fifteen different types of variables have been defined. Then we transfer them into [0.1] interval, or such a transformation is applied to the individual (univariate) distance. The overall distance is calculated through Manhattan approach which makes possible to agglomerate distances on individual axes.

• Details
• Contents