Performance analysis of a proposed ant-based clustering algorithm

Maria Teresinha Arns Steiner, Rosangela Villwock


In the Ant-Based Clustering Algorithm, patterns are spread throughout a grid and each ant is assigned a pattern. The ants are responsible for picking, transporting and dropping patterns on the grid. After the clustering algorithm converges, cluster recovery is done by using the positions of patterns on the grid. The purpose with this study was to evaluate the performance of the Ant-based Clustering Algorithm Proposed (ACAP) compared to the Ant-based Clustering Algorithm – Modified version (ACAM). The major changes were: replacement of the pattern carried by an ant in case it was not dropped within 100 consecutive iterations, comparing the probability of dropping a pattern at a random position with the probability of dropping this pattern at its current position; evaluate the probability of dropping a pattern at a new position, if the pattern is not dropped at a random position, but at a neighboring position. To assess the performance of the algorithm thus proposed, two real examples were used: ÍRIS and WINE. The results show that the ACAP in this study was better than the ACAM for the two examples.


Data Mining; Pattern Clustering; Metaheuristics

Texto completo: