Comparing K-Means and K-Medoids for Industrial Air Pollution Analysis in Central Java

Air is a fundamental necessity for all living beings, especially humans. However, human activities whether intentional or unintentional can degrade air quality through pollution. This study compares the performance of the K-Means and K-Medoids clustering algorithms in analyzing the air pollution load from the industrial sector in Central Java in 2021. Using a quantitative approach and R Studio software, the analysis focuses on SO₂ and NO₂ pollution data obtained from the official Central Java BPS website. The results indicate that the K-Medoids algorithm with the silhouette method yields the most optimal clustering performance, with the lowest Davies-Bouldin Index (DBI) value of 0.6201437 and 10 distinct clusters. Notably, Cluster 1 comprises districts with the highest industrial air pollution burden such as Banjarnegara Regency, which recorded 14,472 industries and NO₂ and SO₂ concentrations of 20 μg/m³ and 6 μg/m³, respectively. These findings demonstrate that clustering algorithms not only help reveal spatial pollution patterns but also provide critical insights for prioritizing targeted mitigation efforts and informing environmental policy-making in industrially active regions.