Automatic document clustering of concept hypergraph decompositions

This paper presents an approach to classify/cluster the web documents by decompositions of hypergraphs. The various levels of co-occurring frequent terms, called association rules (undirected rules), of documents form a hypergraph. Clustering methods is then applied to analyze such hypergraphs; a simple and fast clustering algorithm is used to decomposing hypergraph into connected components. Each connected component represents a primitive concept within the given documents. The documents will then be classified/clustered by such primitive concepts.