A Robust Clustering via Swarm Intelligence

摘要: A very promising approach to reach a robust partitioning is use ensemble-based learning. In this way, the classification/clustering task more reliable, because classifiers/clusterers in ensemble cover faults of each other. The common policy clustering based learning generate set primary partitionings that are different from These could be generated by algorithm with initializations. It popular filter some these partitionings, i.e. subset produced selected for final ensemble. selection phase done diverse consensus function finally aggregates into called also partitioning. Another alternative fusion come naturally sources. On other hand, swarm intelligence new topic where simple agents work such way complex behavior can emerged. necessary diversity achieved inherent randomness algorithms. paper we introduce method on ant colony algorithm. Indeed needs vitally and algorithms inherently involved randomness. Ant powerful metaheuristics concept intelligence. Different runnings dataset result number partitionings. Considering results totally as space employ aggregate them From another perspective, have many parameters. Effectiveness methods questionable they depend test dataset, parameters should tuned obtain desirable result. But how define real does not clear. proposed framework lets free changed, compensates non-optimality power. Experimental real-world datasets presented demonstrate effectiveness generating partitioning..