Hierarchical Clustering Algorithm for Analyzing Risk of Earthquake on Sumatra Island

Abstract

Earthquakes are vibrations produced due to the sudden release of energy from beneath the earth's surface, creating seismic waves. As a part of Indonesia region, Sumatra Island is known for its high level of seismicity, and one of the major earthquakes that caused a tsunami in Aceh occurred in 2004. This study aims to explore the clustering analysis and the hierarchical algorithm of an earthquake on Sumatra Island. This incidence is unpredictable since it occurs in an unexpected location, time, and magnitude. Therefore, to reduce earthquake risk, clustering analysis was carried out in the suspected region. This method includes agglomerative nesting (Agnes) and divisive analysis (Diana) algorithms. They were used in this research due to their effectiveness in grouping objects based on the closest distance or similarities using Euclid's metric. The optimum number of clusters was determined by the silhouette coefficient. The comparison of the cophenetic correlation coefficients in agglomerative nesting gave the conclusion that Ward linkage is the best method with a value of 0.8042. This showed that the solution generated from the clustering process with Ward linkage is quite good. Based on the silhouette coefficient, the Diana algorithm gave better result than the Agnes algorithm for clustering Sumatra earthquake data. The objects of clusters 1 and 2 respectively indicated the occurrence of an earthquake with a high and small risk. The first cluster has larger member than the second, making it susceptible to high earthquake risk.