Improvement of Supervised Isomap Algorithm and Its Application to Visualization and Categorization of Web Chinese Text

It is important to reduce the dimensionality of features in Web Chinese text categorization. Isomap algorithm is an unsupervised manifold learning technique. SIIsomap algorithm, an extension of Isomap to supervised feature extraction, is proposed in this paper. It uses adding constant method and a direct embedding technique of Isomap algorithm for testing data to make the embedding more reasonable and easier. SIIsomap algorithm is applied to visualization and classification experiments of Web Chinese text as a feature extraction method. In contrast with existed methods, it gets better visualization and classification effects and illustrates the effectiveness of our method.