Research On The Classification Of Web-text Based On Support Vector Machines

In this paper, the support vector machines was used to categorize the text-oriented Web content and identify it.there has the complexity of large dimension vector representation and the comparison of the content on the apparent density.besides,there are many relatively strong correlation on it.The traditional skill has poor text classification techniques, Support vector machine is based on the statistical theory of VC dimension and the structure of the minimum principle algorithm. Web text feature in the density and the relationship between factors such as it has little effect on the support vector machine.Support vector machine its own unique processing 'the curse of dimensionality' problem solving skills in dealing with complex Web text shown better in the high-dimensional features. This algorithm is mainly used as the basis of support vector for the angle of the text on the Web text classification research.The main work of this paper is as follows:(1)Analysised the incremental SVM learning algorithm, and based on it, an improved incremental learning with boundary constraints support vector machine algorithm has been proposed. The incremental learning algorithm overcomed the no-comprehensive consideration of factors for the traditional support vector,Proposed boundary constraints, rationally and effectively increased the number of support vector in late. In the basic guarantee for the case of training speed and improve the actual classification accuracy.(2)Analysised the DDAG-SVMS multi-class support vector machine algorithm, and based it on, put forward an improved DDAG-SVMS algorithm. The algorithm t optimize the combination of the original classifier in the he traditional classification algorithms, Improved the structure of the classification tree of the original classifier overcomed the irrational factors, and ultimately improved the actual classification accuracy.(3)Through the experiment, demonstrated that the existing support vector machine technology in the Web text classification and the corresponding efficiency of the improved algorithm.