Your Cart

Your cart is empty

Home  >  Volume 36 (no2)

A COMPARATIVE STUDY OF K-MEANS AND K-MEDOIDS CLUSTERING METHODS. by OSEMWENKHAE J.E.,EKHATOR O.F., and IDUSERI A.(pages 169-176)
Sale price: $5.00
Quantity:

ABSTRACT

The aim of this work is to provide a formal and organized study of the effect of the nature of data and cluster structure on the performance of K-means and K-medoids clustering methods. 

A cluster validation method called Silhouette analysis is used to assess the quality of cluster partitions created by both methods. An illustration on how Silhouette analysis could be used to determine the optimal number of clusters in a data set is presented. Results obtained reveal that the performance of K-means is at its peak with data in which clusters are of relatively uniform sizes while the K-medoids method tends to perform better than K-means when the input data have varied cluster sizes.

Keywords: Cluster Analysis, Cluster Validation,  Distance Functions, K-means, K-medoids, Silhouette Analysis

Close
Loading...