Analýza a klasifikace vícerozměrných shlukových struktur

Abstract

Cluster analysis is generally used to solve many problems such as natural sciences, digital data processing or in the various spheres of daily life. This thesis describes the design of new method of analysis and classification of cluster structures which is based on the using of stochastic standards using information about the properties of the cluster. The aim of development of this method is the classification of clusters of multidimensional structures which is resistant to defects such as defect of the structure and also defects arising due to influence of input data preprocessing. The suggested method uses the basic characteristics of exploratory analysis and its effectiveness is verified by comparative measurements using k-means method. K-means is one of the most widely used method of cluster analysis. The suggested method brings a completely new perspective to cluster analysis within the multidimensional structures.