Hotspot Clustering in Bangka Belitung Islands Province Using Agglomerative Hierarchical Clustering Algorithm

Abstract

Forest and land fires are common disasters in Indonesia, especially in areas with vast forests and lands, such as the Province of Bangka Belitung Islands. Hotspot data, indicating hightemperature areas, can be used to predict the risk of fires. However, not all hotspots indicate actual fires, making it essential to cluster and analyze hotspot data to identify areas with a high likelihood of fires. This study aims to cluster hotspot data in the Bangka Belitung Islands using

the Agglomerative Hierarchical Clustering (AHC) algorithm. The dataset, obtained from NASA's MODIS satellite for the year 2019, consists of 843 records with six attributes, including Brightness, Confidence, and Fire Radiative Power (FRP). Data preprocessing involves cleaning and normalizing the dataset. The AHC algorithm was applied to group hotspots based on their similarity, and Euclidean distance was used to calculate proximity between points. The resulting clusters were evaluated using the Silhouette Coefficient to determine the optimal number of clusters. The results provide insight into areas with varying fire risks, enabling early detection and targeted fire prevention efforts. This study demonstrates the effectiveness of AHC in clustering hotspot data, contributing to better decision-making in forest fire management.