ASSESSING THE EFFECTIVENESS OF CLUSTERING ALGORITHMS IN IDENTIFYING SALINITY DISTRIBUTIONS

Öz

In this study, the performances of various clustering algorithms were compared on salinity data to evaluate their effectiveness in classifying complex spatial patterns. The clustering methods applied included KMeans, Agglomerative Clustering, DBSCAN, MeanShift, Birch, MiniBatch KMeans, and Spectral Clustering. The silhouette score was used as the primary evaluation metric. According to the results, the MeanShift algorithm achieved the best performance with a silhouette score of 0.79, while KMeans and MiniBatch KMeans showed moderate success with scores of 0.38. Agglomerative Clustering, Birch, and DBSCAN yielded silhouette scores of 0.34, 0.31, and 0.28, respectively, whereas Spectral Clustering exhibited the poorest performance with a negative score of -0.35. These findings highlight that density-adaptive methods like MeanShift are particularly effective for analyzing heterogeneous and continuous oceanographic data. The sea water salinity dataset used in this study was obtained from the NOAA World Ocean Database (https://www.ncei.noaa.gov/products/world-ocean-database).

Anahtar Kelimeler

• Clustering Analysis
• Artificial Intelligence
• Sea Water Salinity

TUZLULUK DAĞILIMLARININ BELIRLENMESINDE KÜMELEME ALGORITMALARININ ETKINLIĞININ DEĞERLENDIRILMESI

Öz

Bu çalışmada, farklı kümeleme algoritmalarının tuzluluk verisi üzerindeki performansları karşılaştırılarak, karmaşık mekânsal desenleri sınıflandırmadaki başarıları incelenmiştir. Analizlerde KMeans, Agglomerative Clustering, DBSCAN, MeanShift, Birch, MiniBatch KMeans ve Spectral Clustering algoritmaları kullanılmıştır. Performans değerlendirmesi için siluet skoru temel ölçüt olarak kullanılmıştır. Elde edilen sonuçlara göre, MeanShift algoritması 0.79 siluet skoru ile en iyi sonucu verirken, KMeans ve MiniBatch KMeans algoritmaları 0.38 skoru ile orta düzeyde başarı göstermiştir. Agglomerative Clustering 0.34, Birch 0.31, DBSCAN 0.28 skoru elde etmiş, Spectral Clustering ise -0.35 skoruyla en düşük performansı sergilemiştir. Sonuçlar, özellikle heterojen ve sürekli değişim gösteren okyanus verilerinde yoğunluk adaptif yöntemlerin (örneğin MeanShift) üstün performans sunduğunu göstermektedir. Çalışmada kullanılan deniz suyu tuzluluk verisi, NOAA World Ocean Database’den temin edilmiştir (https://www.ncei.noaa.gov/products/world-ocean-database).

Anahtar Kelimeler

• Kümeleme Analizi
• Yapay Zeka
• Deniz Suyu Tuzluluğu

Kaynakça

1. Jain, A. K., & Dubes, R. C. (1988). Algorithms for clustering data. Prentice Hall.
2. Xu, R., & Wunsch, D. (2005). Survey of clustering algorithms. IEEE Transactions on Neural Networks, 16(3), 645–678.
3. Maze, G., Mercier, H., Fablet, R., Tandeo, P., Radcenco, M. L., Lenca, P., ... & Le Goff, C. (2017). Coherent heat patterns revealed by unsupervised classification of Argo temperature profiles in the North Atlantic Ocean. Progress in Oceanography, 151, 275–292.
4. Ahmed, M., Seraj, R., & Islam, S. M. S. (2020). The k-means algorithm: A comprehensive survey and performance evaluation. Electronics, 9(8), 1295.
5. Chong, B. (2021). K-means clustering algorithm: a brief review. Academic Journal of Computing & Information Science, 4(5), 37–40.
6. Arbelaitz, O., Gurrutxaga, I., Muguerza, J., Pérez, J. M., & Perona, I. (2013). An extensive comparative study of cluster validity indices. Pattern Recognition, 46(1), 243–256.
7. Ranzinger, M., Heinrich, G., Kautz, J., & Molchanov, P. (2024). Am-radio: Agglomerative vision foundation model reduce all domains into one. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 12490–12500).
8. Singh, H. V., Girdhar, A., & Dahiya, S. (2022, May). A literature survey based on DBSCAN algorithms. In 2022 6th International Conference on Intelligent Computing and Control Systems (ICICCS) (pp. 751–758). IEEE.

9. Aidilof, H. A. K., Rosnita, L., Kurniawati, K., & Ikhwani, M. (2025). Clustering of data monitoring water quality using mean-shift clustering method. Journal of Computer Science, Information Technology and Telecommunication Engineering, 6(1).
10. Wahyuningrum, T., Khomsah, S., Suyanto, S., Meliana, S., Yunanto, P. E., & Al Maki, W. F. (2021, December). Improving clustering method performance using K-means, mini batch K-means, BIRCH and spectral. In 2021 4th International Seminar on Research of Information Technology and Intelligent Systems (ISRITI) (pp. 206–210). IEEE.
11. Hanji, S., & Hanji, S. (2023, January). Towards performance overview of mini batch K-means and K-means: Case of four-wheeler market segmentation. In International Conference on Smart Trends in Computing and Communications (pp. 801–813). Singapore: Springer Nature Singapore.
12. Berahmand, K., Mohammadi, M., Faroughi, A., & Mohammadiani, R. P. (2022). A novel method of spectral clustering in attributed networks by constructing parameter-free affinity matrix. Cluster Computing, 25(2), 869–888.
13. Topaloğlu, F. (2024). Saldırı tespit sistemlerinde K-Means algoritması ve Silhouette metriği ile optimum küme sayısının belirlenmesi. Bilişim Teknolojileri Dergisi, 17(2), 71–79.
14. NOAA National Centers for Environmental Information (NCEI). (2025). World Ocean Database. National Oceanic and Atmospheric Administration. Retrieved from https://www.ncei.noaa.gov/products/world-ocean-database