Performance of K-Means Algorithm for Ground Acceleration Clustering

Abstract

Indonesia is one of the most seismically active regions in the world due to the convergence of the Indo-Australian, Eurasian, and Pacific tectonic plates. This condition exposes the country to frequent earthquakes with varying magnitudes and intensities that may cause severe structural damage and pose risks to human safety. Ground acceleration, particularly Peak Ground Acceleration (PGA), is a key parameter for evaluating earthquake impacts and is strongly influenced by geological conditions, hypocentral depth, and epicentral distance. However, the complexity and large volume of ground acceleration data often hinder manual interpretation. This study applies the K-Means clustering algorithm to classify ground acceleration data obtained from seismic records at several observation points. Prior to clustering, data preprocessing was performed through data cleaning and min–max normalization to ensure quality and comparability across variables. The optimal number of clusters was determined using the Elbow method and Silhouette Score. The results reveal distinct distribution patterns of ground acceleration, which are closely related to local seismic conditions. These findings are expected to contribute to the development of preliminary ground acceleration zonation, providing valuable insights for earthquake hazard mapping and risk mitigation efforts in Indonesia.