Comparison of Distance Function to Performance of K-Medoids Algorithm for Clustering

Abstract

The clustering task aims to assign a cluster for each observation data in such a way that observations data within each cluster are more homogeneous to one another than with those in the other groups. Its wide applications in many research fields have motivated many researchers to propose a plethora of clustering algorithms. K-medoids are a prominent clustering algorithm as an improvement of the predecessor, K-Means algorithm. Despite its widely used and less sensitive to noises and outliers, the performance of K-medoids clustering algorithm is affected by the distance function. This paper presents experimentation findings to compare the performance of K-medoids clustering algorithm using Euclidean, Manhattan and Chebyshev distance functions. In this study the K-medoids algorithm was tested using the village status dataset from Gorontalo Province, Indonesia. Execution time and Davies Bouldin Index were used as performance metrics of the clustering algorithm. Experiment results showed that methods of Manhattan distance and Euclidean distance with the Index Davies value of 0.050.