Answers are grounded in this paper and cite their sources.
PT Eka Cipta Mandiri grouped air cargo destinations manually based on subjective estimates, which led to suboptimal cargo space allocation and route pricing. This study applied K-Means clustering, optimized with the Elbow Method and validated with the Silhouette Coefficient, to profile 46 destinations from 2020 to 2025 based on total cargo volume (Kg) and average annual flight frequency. Of 276 annual transaction records, 36 inactive records (zero volume and frequency) were removed, leaving 240 valid records that were aggregated into 46 destinations and scaled with Min-Max normalization. The Within-Cluster Sum of Squares (WCSS) evaluated from K=1 to K=10 showed an elbow at K=3, with a WCSS of 0.8412 and a Silhouette Coefficient of 0.612. Initial centroids were selected from real data points: DJJ (C2), SUB (C1), and AAP (C0). Euclidean distance computations converged at the fourth iteration and divided the destinations into three clusters: Cluster 2 (high/main routes) with 6 destinations (AMQ, BDJ, BTH, DJJ, KNO, UPG), Cluster 1 (medium/growth routes) with 17 destinations, and Cluster 0 (low/niche routes) with 23 destinations. The model was implemented in a web application built with Python, Flask, and SQLite. Testing showed results identical to manual Excel calculations, all Black Box scenarios matching specifications, and successful execution of four independent paths in White Box testing
No takes yet. Share an insight, caveat, or question.
Fadillah et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: