Performance Comparison of Clustering Algorithms to Handle Grouping of City Locations

The study and application of clustering algorithms in real-world problems have become popular in the last years. As a result, we can find in the literature a crescent number of papers with different clustering algorithms and application in several real-world problems. In this paper, we perform an analysis of three clustering algorithms to handle the task of grouping cities aiming to find centroids to serve as locations for designing optical networks. We assessed the K-means, Fuzzy C-means, and Particle Swarm Clustering algorithms. The clustering of city locations is a preprocessing part of an optical network planning tool. We measure the performance in this problem using the Silhouette index as the evaluation metric. The results obtained show a difference in performance between the three algorithms, and also confirms their applicability in this type of problem.