Improving Support Vector Data Description for Document Clustering

Document clustering has received a lot of attention due to its wide application in many fields. To effectively deal with this problem, a new document clustering algorithm is proposed by using marginal fisher analysis (MFA) and improved support vector data description (SVDD) algorithms in this paper. The high-dimensional document data are first mapped into lower-dimensional feature space with MFA. the improved SVDD is then applied to cluster the documents into different classes in the reduced feature space. Experimental results on two document databases demonstrate the effectiveness of the proposed algorithm.