|
|
Cluster analysis as a tool for classification of objects
Budilová, Šárka ; Löster, Tomáš (advisor) ; Šulc, Zdeněk (referee)
Cluster analysis is a popular method of multivariate statistics. Based on mutual similarities between objects this method is able to classify and divide objects into several groups or clusters. The results of the clustering can be different by using different methods, measures of distance and procedures. The main aim of this thesis is to compare the results of several methods of cluster analysis with the known classification of classes from the original data file. In total, there are 15 data files, which were analyzed and each of them contained known information about the right allocation of objects in groups. The success of clustering of each method was calculated by comparing the known classification of classes and resulted clusters. In addition to the comparison of individual methods of cluster analysis was compared the impact of standardization and correlation to the success of each method. To reflect the distance betweeen the objects within each clusters, squared Euclidean distance was used. The results of this thesis point out that better success of clustering were achieved in the case of correlated variables in data file. The succes of clustering was higher about 2 percent points than in the case when correlated variables were deleted from data set. The methods divided 69,8 % objects before standardization and 70,8 % objects after standardization. The results also show a large importance of standardization in the case of Ward´s method. After standardization this method rank the most objects into correct classification classes and were more succesful, about nine percent points. In the case of correlated variables is the succes of the method 76,4 %. Standardization positively influences also centroid method and the method of farthest neighbour. Median method, nearest neighbour method and the method of average linkage achieve higher success of clustering in the case of original, nonstandardized variables (uneven variables).
|
|
|
Discriminant and cluster analysis as a tool for classification of objects
Rynešová, Pavlína ; Löster, Tomáš (advisor) ; Řezanková, Hana (referee)
Cluster and discriminant analysis belong to basic classification methods. Using cluster analysis can be a disordered group of objects organized into several internally homogeneous classes or clusters. Discriminant analysis creates knowledge based on the jurisdiction of existing classes classification rule, which can be then used for classifying units with an unknown group membership. The aim of this thesis is a comparison of discriminant analysis and different methods of cluster analysis. To reflect the distances between objects within each cluster, squeared Euclidean and Mahalanobis distances are used. In total, there are 28 datasets analyzed in this thesis. In case of leaving correlated variables in the set and applying squared Euclidean distance, Ward´s method classified objects into clusters the most successfully (42,0 %). After changing metrics on the Mahalanobis distance, the most successful method has become the furthest neighbor method (37,5 %). After removing highly correlated variables and applying methods with Euclidean metric, Ward´s method was again the most successful in classification of objects (42,0%). From the result implies that cluster analysis is more precise when excluding correlated variables than when leaving them in a dataset. The average result of discriminant analysis for data with correlated variables and also without correlated variables is 88,7 %.
|
guest ::
