Application of the K-Medoids Algorithm to Group Obesity Status

Abstract

Obesity is a global health problem that is increasingly urgent to be addressed. The World Health Organization (WHO) notes that the incidence of obesity has tripled since 1975, with more than 1.9 billion adults being overweight, and of these, more than 650 million people diagnosed with obesity. This research aims to apply the K-Medoids algorithm in grouping obesity status data. Clustering was performed using RapidMiner with various configurations of the number of clusters, and results were evaluated using several cluster validation metrics including the Davies-Bouldin Index (DBI). The results showed that three clusters (k = 3) were the most optimal in grouping obesity data, with a DBI value of 0.071, compared to two clusters (k = 2) which had a DBI value of 0.101. A lower DBI value indicates that the clusters formed are more compact and well separated, indicating better performance in grouping obesity status data.