Evaluating the Impact of Distance Metrics on K-means Clustering Performance with Geospatial Data
Küçük Resim Yok
Tarih
2026
Yazarlar
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Springer Science and Business Media Deutschland GmbH
Erişim Hakkı
info:eu-repo/semantics/closedAccess
Özet
Clustering algorithms are fundamental in data mining, machine learning, and statistical analysis, with K-means being one of the most popular due to its simplicity and efficiency. This study examines the impact of using different distance metrics (Euclidean and Vincenty) on the performance of the K-means clustering algorithm when applied to geographical data. While Euclidean distance is widely used for its straightforward computation, it may not be ideal for clustering geographical locations due to the Earth’s curvature. Vincenty’s formula, which accounts for the Earth’s ellipsoid shape, offers a potentially more accurate alternative. We conducted experiments using a dataset of 69,902 geographical points, evaluating the clustering performance based on Sum of Squared Errors (SSE), Davies-Bouldin Index (DBI), and Silhouette Index. Our results indicate that although Vincenty’s formula theoretically provides more precise distance measurements depending on the earth shape, the practical differences in clustering performance between Euclidean and Vincenty distances are minimal. Statistical analysis using the Mann–Whitney U test demonstrated that the observed differences lacked statistical significance. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.
Açıklama
3rd International Conference on Data, Electronics and Computing, ICDEC 2024 -- 18 September 2024 through 20 September 2024 -- Kayseri -- 344109
Anahtar Kelimeler
Clustering, Euclidean, K-means, Machine learning, Vincenty
Kaynak
Lecture Notes in Networks and Systems
WoS Q Değeri
Scopus Q Değeri
Q4
Cilt
1455 LNNS
