Klasterska analiza

Ungaro, Tea (2016) Klasterska analiza. Diploma thesis, Faculty of Science > Department of Mathematics.

[img]
Preview
PDF
Language: Croatian

Download (1MB) | Preview

Abstract

This paper describes cluster analysis and its application to the base of diabetics divided according to degree of the disease. Cluster analysis involves grouping elements into clusters so the elements within one cluster are the most similar. For clustering we must first define the distance (similarity) between two elements and then define the grouping algorithm. There are hierarchical and nonhierarchical clustering. In described example there were 13 variables measured from which was expected that patients will be grouped into four clusters (healthy, type I diabetes, type II diabetes, prediabetes). Hierarchical clustering was made with the Euclidian distance and the average method, the Gower distance and the Ward method and the Manhattan distance and the maximum method. Also, nonhierarchical clustering was made with k-means algorithm for k=4 and 3.

Item Type: Thesis (Diploma thesis)
Supervisor: Jazbec, Anamarija
Date: 2016
Number of Pages: 38
Subjects: NATURAL SCIENCES > Mathematics
Divisions: Faculty of Science > Department of Mathematics
Depositing User: Iva Prah
Date Deposited: 17 Nov 2016 11:05
Last Modified: 17 Nov 2016 11:05
URI: http://digre.pmf.unizg.hr/id/eprint/5298

Actions (login required)

View Item View Item