Clustering ensemble framework via ant colony

Abstract

Ensemble-based learning is a very promising option to reach a robust partition. Due to covering the faults of each other, the classifiers existing in the ensemble can do the classification task jointly more reliable than each of them. Generating a set of primary partitions that are different from each other, and then aggregation the partitions via a consensus function to generate the final partition, is the common policy of ensembles. Another alternative in the ensemble learning is to turn to fusion of different data from originally different sources. Swarm intelligence is also a new topic where the simple agents work in such a way that a complex behavior can be emerged. Ant colony algorithm is a powerful example of swarm intelligence. In this paper we introduce a new ensemble learning based on the ant colony clustering algorithm. Experimental results on some real-world datasets are presented to demonstrate the effectiveness of the proposed method in generating the final partition. © 2011 Springer-Verlag.