PERBANDINGAN ALGORITMA SINGLE LINKAGE, COMPLETE LINKAGE DAN AVERAGE LINKAGE DALAM PENGELOMPOKAN PROVINSI DI INDONESIA BERDASARKAN JENIS KEJADIAN BENCANA ALAM

Abstract

The number of natural disaster events in Indonesia in 2021 increased almost twofold compared to the previous three years. It is necessary to group areas based on their similar characteristics using cluster analysis for mitigation efforts purposes. This study aims to (1) analyze the best algorithm amongst the Single Linkage, Complete Linkage, and Average Linkage, and (2) group provinces in Indonesia based on the type of natural disaster using the best algorithm. These three algorithms are compared using Silhouette coefficient values. Data on the number of natural disasters in each province in Indonesia, sourced from Statistics Indonesia (BPS), were used. The Single Linkage, Complete Linkage, and Average Linkage produce Silhouette coefficient values of 0.65; 0.65; and 0.68, respectively, with the optimal number of clusters of four, three, and three, respectively. The Average Linkage produces the maximum Silhouette coefficient value, thus the best algorithm in this research. Clusters resulting from this algorithm show that 29 provinces have a moderate tsunami intensity and other disasters are low; two provinces (North Sumatra, East Java) have high intensity earthquakes, tsunamis and volcanic eruptions, while other disasters are moderate; three provinces (West Java, Central Java, East Nusa Tenggara) have relatively low tsunami intensities, moderate earthquakes and volcanic eruptions, while other disasters are high.

Keywords: Single Linkage, Complete Linkage, Average Linkage, natural disaster, Silhouette coefficient.