Web categorization using hybrid algorithms

Abstract

Obtaining information from the Web is becoming a very much important issue nowadays. The traditional text categorization algorithm is not sufficient for web categorization. In this paper we discuss the process in Web categorization, and proposed a new information gain measure for feature selections and term weighting. We also discussed three linear classifiers. Then we propose a novel hyperlink based classifier. It uses the characteristics of the Web graph. Experimental comparisons of these algorithms show that our approach is more appropriate than traditional information retrieval methods in Web categorization.