An ACO-Based Clustering Algorithm

MACOC: A Medoid-Based ACO Clustering Algorithm

“…The main different, apart of the algorithm centroid nature, is that ACOC uses a pheromone matrix from the data instances to the centroidlabels, while our algorithm use a pheromone matrix between all the data to remember the previous medoid assignation. The parameters of ACOC and MA-COC algorithms have been set in a similar way to the original work [10]: the number on ants is 10, the number of elitism is 1, the exploration probability is 0.0001, the initial pheromone values follow an uniform distribution [0.7, 0.8], β = 2.0, ρ = 0.1, and the maximum number of iterations is 1000. The only difference is that the MACOC initial pheromone values have been set as 1 n (where n is the number of clusters).…”

“…From other bio-inspired perspectives, ACO algorithms have also produced promising results. Kao and Cheng [10] introduced a centroid-based ACO clustering algorithm; França et al [7] introduce a bi-clustering algorithm; and Ashok and Messinger focused their work on graph-based clustering [1]; several other approaches are discussed in [9].…”

“…As an ACO algorithm, MACOC algorithm is based on ACOC algorithm [10]. They have a similar search graph, where the ants try to define the optimal cluster assignment for each of the instances (data points).…”

“…It has been inspired by a previous algorithm, called ACO-based Clustering algorithm (ACOC), proposed by Kao and Cheng [10]. ACOC is a centroid-based clustering algorithm, which tries to optimize the centroid (central point) position of each cluster.…”

A charged system search approach for data clustering

“…The simulation results indicate that the proposed algorithm has provided better results in terms of quality of solutions. Kao et al [22] have proposed an ACO-based algorithm for clustering and named it ACOC. The performance of ACOC algorithm is compared with K -means and Shelokar ACO algorithm in which ACOC has given better results.…”