Research On Text Classification Based On Support Vector Machine With Mixture Of Kernels

With the development of computer technology, people’s ability of gathering andstoring data has greatly improved. Not only in the scientific research area but also invarious fields of social life, a large amount of data has been accumulated. Dataanalyzing and data mining that using machine learning methods contributed to thegeneration of support vector machine classification technology. Since Vapnikproposed the Support Vector Machine based on Statistical Learning Theory and kerneltrick in1990s, kernel method based machine learning algorithm has been developedrapidly, which becomes one of the hot points of machine learning and artificialintelligence fields, and be widely used in biology information technology, imageprocessing, text classification and so on.Further research on support vector machine is of great significance for both thedevelopment and improvement of the kernel theory and the expansion of application.The kernel function is the important way to realize non-linear mapping, which is theessence of Support Vector Machines with such wide application. This paper is todiscuss the structure, properties and applications of hybrid kernel function. Thesignificance of studying hybrid kernel is to enhance the applicability of SVM so as togive support to pattern analysis, artificial intelligence and machine learning. On theother hand, kernel function method is just at the initial stage, its potential has not beenfully excavated.As the data mining technology is developing so fast, today’s text categorizationtechnology can improve the status of text message disorder and improve searchquality, and get access to text message efficiently. Therefore, automatic textclassification technology has attracted more attention. Text categorization technologybased on machine learning has a good effect, and there are a variety of classificationalgorithms, such as: KNN algorithm, naive Bayes algorithm, decision tree algorithmand support vector machine algorithm. This paper applied support vector machine algorithm based on hybrid kernel totext classification technology. At first, the legitimacy, nature, algorithm of new hybridkernel were discussed, and the paper describe the WEB text classification methodsteps: text preprocessing, feature reduction, text feature representation method, andthen constructs a model of WEB text classification based on support vector machinestructure, through the simulation experiment shows that the new hybrid kernelfunction is better than that of single kernel as well as the commonly used hybridkernel in accuracy and efficiency.