Perbandingan K-Means, DBSCAN, Agglomerative pada Clustering Gempa Bumi Sumbagsel 2015–2025 Berbasis USGS

Abstract

Southern Sumatra is one of the most seismically active regions in Indonesia, situated along the Great Sumatran Fault and the Indo-Australian subduction zone, making systematic spatial analysis of earthquake data essential for effective disaster mitigation. This study conducts a comparative analysis of three clustering algorithms, namely K-Means, DBSCAN, and Agglomerative Clustering, applied to earthquake data in the southern Sumatra region for the period 2015–2025 sourced from the United States Geological Survey (USGS) catalog. A total of 2,483 earthquake events were analyzed after preprocessing, with key attributes including geographic coordinates (latitude and longitude) and hypocenter depth. Cluster quality was evaluated using Silhouette Score and Davies-Bouldin Index (DBI). Results show that K-Means achieved the best quantitative performance with a Silhouette Score of 0.4204 and DBI of 0.8715, followed by Agglomerative Clustering with a Silhouette Score of 0.4047 and DBI of 0.8789. DBSCAN could not be evaluated quantitatively due to noise point classification, yet effectively captured complex spatial patterns aligned with the subduction zone structure. All three methods consistently identified seismic activity concentrated along the southern Sumatra subduction zone, dominated by shallow to intermediate-depth earthquakes. These findings provide data-driven recommendations to support earthquake disaster mitigation planning in the southern Sumatra region