A CLUSTERING ALGORITHM BASED ON SWARM INTELLIGENCE FOR WEB DOCUMENT

Swarm intelligence due to its flexibility, robustness and self-organization has been applied in a variety of areas. A clustering algorithm based on swarm intelligence （CSI） for web documents is proposed. Firstly, web documents, which are denoted by vector space model with reduced document feature set, are randomly projected on a plane. Then, clustering analysis is conducted by a clustering method derived from a basic model interpreting ant colony organization of cemeteries. The artificial ants perform random walks on the plane and pick up or drop projected data items with the probability which is converted from swarm similarity within a local region by probability conversion function. Clusters are visually formed on the plane by ant colony collective actions in the absence of central controls. Finally, the clustering results are collected from the plane by a recursive algorithm. Each clustering center is labeled by the most weighted feature. A hybrid clustering algorithm CSIM is also proposed by combining the CSI with the k -means algorithm. CSIM inherits the prominent properties of both swarm intelligence and k -means. It also offsets the weakness of those two techniques. The experiment results and the comparison with other document clustering methods show that this web document clustering algorithm based on swarm intelligence has good clustering performance. The web documents focusing on a subject are rather completely and exactly clustering together.