Evaluating the success of cluster analysis methods

Cluster analysis is one of the classification methods of multivariate statistical analysis. The task of this analysis is to classify the objects into clusters so that objects inside these clusters are as similar as possible. The aim of this study is to evaluate the success of the classification of objects using six hierarchical cluster analysis methods. To reflect the distance between the objects, are used squared Euclidean and Mahalanobis distances. The success methods are evaluated through the information, which cluster the object belongs to, and this information is already contained in the data files. This thesis pointed out that the Ward's method is one of the most successful hierarchical method in a classification of objects into clusters. This method has been more successful in sorting objects than the other hierarchical methods, both in the case of leaving the correlated variables in the data file as well as removing them. The results of this work show that the highest success of classification objects into clusters is when the data set is cleaned of correlated variables. If the data file is not cleaned, the methods reach better results when the distance between objects is measured by Euclidean metric