Application Research Of Web Mining Based On Radial Basis Function Neural Network

As the rapid progress of the Internet, the amount of information in the WEB becomes larger and larger, and the WWW has become a global, .huge., distributing and shared information resources. Additionally, this situation is developing with high speed. The WEB resource has no common data model, query language and common structure because it is distributed and always changed. As a result, users are always faced with some problems like "information maze " or "information over loading". So it is urgent for us to find a way to help users to find information that they really want. Although some technologies, such as, powerful search engine, intelligent Agent and WEB query language, are helpful, but they always return some information which is not needed by users. It will influence the precision ratio and the efficiency of the query. So how to identify knowledge from mass data is a crucial problem, and the proper classification on the WEB page is the key to solve this problem.In this thesis, on the base of WEB classification problems, we develop research, how to classify the WEB page automatically according to some subjects and the method to construct the classifier. And we design a WEB classification mining systemmodel based on RBF neural network. In order to judge effectively that which category one WEB page should belong to and find the WEB page that can reflect some subject, we put forward a classifier constructed by Radial Basis Function (RBF) neural network. By using the classifier, we can classify the WEB documents automatically. At the same time, we compare the RBF with the BP neural network through experiments and found that the RBF classifier performance better than BP classifier.We discussed the WEB mining and the object expression of the WEB document at first, introduced learning algorithms of the RBF in detail, and discussed the design ideas and the correlative technologies of the classify engine about some WEB subjects at last. We focused on researching the structure of RBF neural network classifier and its realization processes. We brought forward our view of using dynamic nearest clustering algorithm to study RBF WEB, and did an experiment to test it.