Comparison of K-Means and K-Medoids Clustering Algorithms for Mapping the Spread of Tuberculosis Disease in Central Java Province

Abstract

Along with the development of information technology, data analysis plays an important role in various fields, including in the health sector. This study aims to map the spread of Tuberculosis (TB) in Central Java Province using two non-hierarchical clustering algorithms, namely K-Means and K-Medoids. The data used are TB case data in 2024 from 35 districts/cities. The application of the clustering algorithm is carried out using three types of distances, namely Euclidean, Manhattan, and Minkowski. Evaluation of cluster ing results using the Davies-Bouldin Index validity index. The index values obtained from the K-Means algorithm are the Euclidean distance of 0.7822, Manhattan of 0.7789, and Minkowski of 0.7822. Meanwhile, the K-Medoids algorithm provides index values at Euclidean distances of 0.7803, Manhattan of 0.7950, and Minkowski of 0.7803. Based on these results, the K-Means algorithm with Manhattan distance showed the best performance because it produced the lowest Davies-Bouldin Index value, which reflects a more optimal and valid cluster  structure. In general, the areas included in the high TB case category and need to be the main focus of handling are Banyumas, Brebes, Cilacap, Tegal, and Semarang City. These findings are expected to be the basis for consideration in making more targeted TB control policies.