PERFORMANCE COMPARISON OF K-MEDOIDS AND DENSITY BASED SPATIAL CLUSTERING OF APPLICATION WITH NOISE USING SILHOUETTE COEFFICIENT TEST

Abstract

Cluster analysis is a technique for grouping objects in a database based on their similar characteristics. The grouping results are said to be good if each cluster is homogeneous, and can be validated using the silhouette coefficient test. However, the presence of outliers in the data can affect the grouping results, so methods that are robust to outliers are used, such as K-Medoids and Density-Based Spatial Clustering of Applications with Noise. The purpose of this study is to compare the results and performance of the two methods using the silhouette coefficient test on data on human development indicators in South Sulawesi Province in 2021. The results of the analysis show that K-Medoids produced 2 groups, namely the districts/cities group which has indicators of human development that consist of 21 districts/cities, and the high group, which consists of 3 districts/cities, while Density-Based Spatial Clustering of Application with Noise produces 1 group that has the same characteristics, which consists of 19 districts/cities, and the remaining 5 districts/cities are identified as noise. Based on the silhouette coefficient test, K-Medoids have a greater value than Density-Based Spatial Clustering of Application with Noise, namely 0,635 and 0,544, respectively, so that K-Medoids have better performance.