Research on Web Page Classification Method based on Newton’s Law of Universal Gravitation and HITS Algorithm

Web page classification is one of the most important methods in web mining. In recent years, numerous classifiers are proposed and used for web page classification. Though these classifiers perform well in practice, they don't pay enough attention to the link connections between web pages, and therefore, their classification efficiencies can’t be greatly improved. We propose a web page classification method based on Newton’s Law of Universal Gravitation and HITS (Hypertext-Induced Topic Search) algorithm (WPCM), based on which, we constructs the web page classification system. In this system, web pages are firstly collected and pro-processed; Web pages are represented as the document-term frequency matrix and Modified Discriminant Analysis based on Fisher Criterion and Manifold Learning (MDA) is used to feature extraction; Web pages are classified by WPCM; At last, the performance of WPCM is evaluated by the criterions of Accuracy and F-measure. We use PKU collection, containing 13897 web pages and 11 categories, in our experiments. The comparative experiments with the traditional classifiers, such as SVM and KNN, demonstrate that the proposed system based on the classifier WPCM performs better than SVM and KNN.Inspired by Newton’s Law of Universal Gravitation, we proposed a new web page classification method WPCM, in which the research object is transformed from the object in the universe to the web page and the importance representation is transformed from the mass of the object to the weight of the web page. The above research route can broaden related research horizon and may cause the researcher's interest.