A clustering ensemble learning method based on the ant colony clustering algorithm

Abstract

Ensemble-based learning is a successful approach for robust partitioning. Since the ensemble classifiers cover each other fault, classification is a critical task. Clustering ensemble based learning can also be done using fusion of some primary partitions which derive from naturally different sources. In this study, a novel clustering ensemble learning method inspired from the ant colony clustering algorithm is proposed. Since ensemble methods necessarily rely on diversity, swarm intelligence algorithms, such as ant colony, are can be good options to be applied. Executing this algorithm for several times on a dataset, result in various partitions. Then, a simple partitioning algorithm is exercised to aggregate them into a consensus partitioning. The proposed clustering approach lets the parameters be free to be manipulated, and thanks to the ensemble, non-optimality of the parameters is covered. Experimental results on several real datasets illustrate the efficiency of the proposed method to generate the final partitioning.