An Efficient Spectral Method for Document Cluster Ensemble

Cluster ensemble techniques have been recently shown to be effective in improving the accuracy and stability of single clustering algorithms. A critical problem in cluster ensemble is how to combine multiple clusterers to yield a final superior clustering result. In this paper, we present an efficient spectral graph theory-based ensemble clustering method feasible for large scale applications such as document clustering. Since the EigenValue Decomposition (EVD) of Laplacian is formidable for large document sets, we first transform it to a Singular Value Decomposition (SVD) problem, and then an equivalent EVD is performed. Experiments show that our spectral algorithm yields better clustering results than other cluster ensemble techniques without high computational cost.