A filter-based feature selection method for web page classification

In this study, filter based methods are proposed to select best features for Web page classification problem to improve accuracy and to reduce time required for classification. Web pages are classified by selecting the best features using Sort Orders at Partial Query Expansion algorithms and IDF method. LibSvm, NaiveBayes Multinomial and J48 classifiers are used. The proposed Sort Order algorithms are experimented on the WebKB, Conference, Bank Search and Syskill Webert datasets. The feature selection algorithms are also compared with the two most popular attribute selection methods that are Information Gain and ChiSquare to evaluate their performance. The experimental results of this study showed that, Sort Orders at Partial Query Expansion are good at with feature selection for Web page classification tasks.