COMPARISON OF CLUSTERING EARTHQUAKE PRONE AREA IN SUMATRA ISLAND USING K-MEANS AND SELF-ORGANIZING MAPS

Abstract

An earthquake is a sudden vibration on the earth's surface caused by the shifting of tectonic plates. One region in Indonesia that is particularly prone to earthquakes is Sumatra Island, due to its geographical location at the convergence of two tectonic plates, namely the Indo-Australian plate, which is actively subducting beneath the Eurasian plate. While earthquakes cannot be prevented or avoided, effective disaster mitigation strategies can help minimize the impact. The purpose of this research is to classify earthquake-prone areas on Sumatra Island based on depth and magnitude, allowing for further analysis to determine the characteristics of the clustering results. The study employs two clustering methods to analyze earthquake data from 1973 to 2024: the K-means and Self-Organizing Maps (SOM) algorithm. K-means algorithm is preferred for its simplicity and efficiency in handling large datasets, and suitability for numerical earthquake data analysis. Conversely, the SOM algorithm offers more stable clustering results and preserves the topological structure of the data, making it advantageous for exploring spatial patterns. The research findings indicate that the K-means algorithm provides better grouping, achieving a Silhouette Coefficient of 0.53, compared to 0.47 for the SOM algorithm. The K-means clustering resulted in two clusters: Cluster 1 contains 1,213 members and is characterized by shallow depths (3.9 km-41 km) and larger magnitudes (5 - 8.92 ), indicating a higher risk level. In contrast, Cluster 2 includes 412 members and represents areas with greater depths (40.8 km-70 km) and smaller magnitudes (5 - 6.85 ), corresponding to a lower risk level. This research aims to support the government in its earthquake disaster mitigation efforts, especially on Sumatra Island.