Classification of websites as sets of feature vectors

The world wide web is the largest source for all kind of information currently available. Due to its enormous size retrievin2 relevant information is a difficult task for which users often rely on directory services. A directory service provides a huge topic tree containing links for each topic. Due to the generality of the topics most links direct to websites or domains. instead of single webpages. For maintaining a directory service, automatic classification of new websites into the topics of the tree would be very beneficial. Therefore, this paper introduces a new approach to website classification that is based on sets of feature vectors. Compared to previous approaches our new method requires no preprocessing, but provides high accuracy in efficient time.