Comparison between Agglomerative Method, Divisif Method, and K-Means Method in Cluster Analysis (by Dewi Rachmatin and Kania Sawitri)

It was announced at the UNPAR National Seminar on Mathematics in 2014


The process of grouping data in cluster analysis can be done with two methods: hierarchy method and non-hierarchy method. Hierarchical methods are clustering methods that form the construction of a hierarchy based on a certain level such as a tree structure. This method is divided into two, namely the agglomerative (concentration) and divisif (deployment) methods. Methods that include agglomerative methods include: Single Linkage Method, Complete Linkage Method, Average Linkage Method, Ward’s Method, Centroid Method and Median Method. The six agglomerative methods (hierarchy) have been discussed in previous research (Rachmatin, 2012). In the hierarchy method the number of groups to be obtained is not yet known, whereas the non-hierarchy method starts by assuming there is a k group first. Methods that include non-hierarchical methods are the k-means method and the fuzzy k-meansmethod. In this article discussed the results of theoretical studies which are comparisons of hierarchy methods (agglomerative methods and divisif methods) with non-hierarchical methods represented by the k-meansmethod. The theoretical study was applied to a data that is air pollution level data (Gunawan et al., 2010; Rachmatin, 2012).

Keywords :

Agglomerative Method, Divisif Method, and K-Means Method.