FUZZY GEOGRAPHICALLY WEIGHTED CLUSTERING WITH OPTIMIZATION ALGORITHMS FOR SOCIAL VULNERABILITY ANALYSIS IN JAVA ISLAND

Fadlurohman, Alwan
Utami, Tiani Wahyu
Amrullah, Setiawan
Roosyidah, Nila Ayu Nur
Dhani, Oktaviana Rahma

BAREKENG: Jurnal Ilmu Matematika dan Terapan
Vol 19 No 3 (2025): BAREKENG: Journal of Mathematics and Its Application

The Social Vulnerability Index (SoVI) measurement assesses social vulnerability. However, the measurement of SoVI can only describe the general conditions without being able to show which factors dominate. Therefore, a clustering approach has been proposed to characterise the dominant social vulnerability factors. Fuzzy Geographically Weighted Clustering (FGWC) is a method that works for this purpose. FGWC is an extension of the Fuzzy C-Means algorithm, which involves geographical influences in calculating membership values. However, the FGWC method is sensitive because the initial initialisation to determine the centroid is randomised, and it will affect the cluster quality. This research uses a metaheuristic approach to overcome the weakness of FGWC by using Particle Swarm Optimisation (PSO) and Artificial Bee Colony (ABC). This study aims to cluster districts/cities in Java Island using the PSO-FGWC and ABC-FGWC methods based on social vulnerability variables and determine the dominant factors of social vulnerability in each region. Optimum cluster selection uses the index of the largest Partition Coefficient (PC) and the smallest Classification Entropy (CE). Clustering social vulnerability in Java Island resulted in the best clustering using the ABC-FGWC method with 5 optimum clusters based on the PC and CE index values of 0.343 and 1.298, respectively. This research found that social vulnerability exists in each region of Java Island. Cluster 1, consisting of 19 districts/cities, is characterized by vulnerabilities in demography and education. Cluster 2, consisting of 33 districts/cities, is characterized by demographic and health vulnerabilities. Cluster 3, which consists of 24 districts/cities, is dominated by education and economic vulnerability factors. Cluster 4, consisting of 14 districts/cities, has the highest social vulnerability characteristics on the unemployment rate and the proportion of house rent. The last one, cluster 5, consists of 29 districts/cities and has a vulnerability problem in the population growth variable.