Web Document Classification Using Naïve Bayes

Abstract

World Wide Web has become a huge collection of documents and the amount of documents available is increasing on a daily basis. How to correctly classify the vast documents into a particular category and locate any document of interest easily has become a challenge researchers have been trying to solve for decades and different researchers have attempted different algorithms using different platform to achieve this aim. In this paper, a University web site was used as a case study and a machine learning workbench called WEKA (Waikato Environment for Knowledge Analysis) which provides a general-purpose environment for automatic classification, regression, clustering and feature selection was used as a machine learning platform. Running Naïve Bayes with 10-fold cross validation on the selected web data gives a 77% correctly classified instances in zero second with relative absolute error of 68.9937%. This shows the ability of Naïve Bayes algorithm to accurately classify vast amount of web document in a short time.