SVM multi-classifier and Web document classification

This paper deals with support vector machine for multi-classification problem and its application to Web page document classification. Several multi-classifier of SVM are mentioned, their construction and computing complexity are compared and analyzed. For special application problems, most of multi-classifier of SVM are limited on the number of class. Web page document classification is a classical multi-classification problem; also the number of samples and the scale of dimension are so large that many classifiers are inefficient, such as multi-layer neural networks, RBF neural networks, k-neighbor, etc. SVM is the first choice for Web page document classification because of its advantage on non-effective of feature dimension scale. This paper focuses on direct design of multi-classifier of SVM and its application to Web page classification. The experiment results illustrate the efficiency of this kind of classifier.