A max min ant system applied to the capacitated clustering problem

• Published in: Proceedings of the 2004 14th IEEE Signal Processing Society Workshop Machine Learning for Signal Processing, 2004 pp. 755 - 764
• Main Authors: de Franca, F.O., Von Zuben, F.J., Nunes de Castro, L.
• Format: Conference Proceeding
• Language: English
• Published: Piscataway NJ IEEE 2004; New York NY
• Subjects: Applied sciences; Cities and towns; Clustering algorithms; Computer science; control theory; systems; Data mining; Data processing. List processing. Character string processing; Electronic mail; Exact sciences and technology; Informatics; Large-scale systems; Memory organisation. Data processing; Mining industry; Pattern recognition; Proposals; Software; Vehicles; Cluster analysis; Performance evaluation; Local search; Insecta; Updating; Information retrieval; Social insect; Formicoidea; Modeling; Location problem; Optimal allocation; Arthropoda; Heuristic method; Minimax method; Hymenoptera; Invertebrata; Aculeata
• ISBN: 0780386084; 0780386086; 9780780386082
• ISSN: 1551-2541; 2378-928X
• DOI: 10.1109/MLSP.2004.1423042

This work introduces a modified max min ant system (MMAS) designed to solve the capacitated clustering problem (CCP). Some improvements on the original MMAS algorithm are proposed, such as the use of a density model on the information heuristic and a local search adapted from the uncapacitated p-medians problem. Also the MMAS ability to deal with large scale instances is improved by means of a new proposal for the pheromone updating rule. Some simulations are performed using instances available from the literature, for benchmarking purposes. As a practical application, given a hypothetical demand proportional to the number of inhabitants of the 186 most populated Brazilian cities, the optimal allocation for a varied number of clustering centers is properly determined by the proposed algorithm, with a superior performance when compared with the original MMAS algorithm