A document clustering algorithm based on improved landmark semidefinite embedding

The document space is generally of high dimensionality, and clustering in such a high dimensional space is often infeasible due to the curse of dimensionality. In this paper, a novel document clustering method which based on improved landmark semidefinite embedding (lSDE) is proposed. Based on the general lSDE, the point selection rules is modified by Max-min distance algorithm, with a view to ensuring the stability of algorithm. By using the improved lSDE, the documents can be projected into a lower dimension kernel space in which redundant information was filtered, and the documents related to the same semantic are close to each other. On this low-dimensional representation, the processed document data was clustered by kernel K-means. Experimental results show that the new clustering algorithm gives better performance than several advanced clustering methods.