A Kind Of Transductive Inference WEB Data Mine Based On Support Vector Machine

With the explosive growth of the Internet, WEB has contained flood of information and rich resources. It is necessary to provide users with tools for efficient knowledge discovery on the WEB, in order to improve the efficiency of information retrieval and information usage on WEB.With the application of data mining to WEB, WEB mining has become one of the most important and flourishing fields of data mining.SVM is a new kind of promising machinel earning algorithm proposed by Vapnik and his group at AT&T Bell laboratory. Because SVM has stronger theoretical foundation and better generalization performance, it becomes the new research hotspot after the research of Artificial Nerve Net and it will push the development in machine learning theory and technology.In this paper, the principle of WEB mining has been introduced at first.The building of WEB text mining system has been discussed in detail, including the architecture of WEB text mining system and function of system module.Secondly,the basc knowledge of the statistical learning theory has been introduced and the SVM based on the theory has been gone deep into discussed. Tirdely,by combining the characteristics of SMV we carry out a method of Progressive Transductive Support Vector Machine called PTSVM, and make out the specific step of practicing and the provement of the algorithm .In the end we give a result of the experiment and detailed analysis.