Web Document Classification by Keywords Using Random Forests

Web directory hierarchy is critical to serve user’s search request. Creating and maintaining such directories without human experts involvement requires good classification of web documents. In this paper, we explore web page classification using keywords from documents as attributes and using the random forest learning methods. Our initially results are promising that the random forests learning method performed better than several other well known learning methods. When the number of topics increased from five to seven, random forests still performed better than other methods even though absolute classification rates decreased.