In this paper, a swarm of agents are created to solve clustering ensemble problem. An agent takes action according to the environmental information, whereas the result of action feed back to the environment. A swarm of agents with well-designed individual model are applied for integrating clusterings of diverse algorithms into a consensus clustering. The agent's individual model is so effective that the complex global objective can be achieved through agents pursuing their private goal. As a result, the consensus clustering is more similar to the "true clustering" than initial clusterings.
展开▼
