Comparison analysis of Euclidean and Gower distance measures on k-medoids cluster

Abstract

K-medoids is a clustering method that uses the distance method to find and classify data that have similarities and inequalities between data. This shows that the determination of the distance measurement method is important because it affects the performance of the k-medoids cluster results. From several studies, it is stated that the Euclidean and Gower methods can be used as measurement methods in clustering with numerical data. This study aims to compare the performance of the k-medoids clustering results on a numerical dataset using the Euclidean and Gower methods. The method used is the Knowledge Discovery in Database (KDD) method. In this study, seven numerical datasets were used and the evaluation of clustering results used silhouette, Dunn, and connectivity values. The Euclidean distance method is superior to the two values of silhouette evaluation and connectivity, it shows that Euclidean has a good data grouping structure, while the Gower is superior to the Dunn value, which shows that the Gower has good cluster separation compared to Euclidean. This study shows that the Euclidean method is superior to the Gower method in the application of the k-medoids algorithm with a numeric dataset.