Effective Text Classification by a Supervised Feature Selection Approach

The high dimensionality of data is a great challenge for effective text classification. Each document in a document corpus contains many irrelevant and noisy information which eventually reduces the efficiency of text classification. Automatic feature selection methods are extremely important to handle the high dimensionality of data for effective text classification. Feature selection in text classification focuses on identifying relevant information without affecting the accuracy of the classifier. Several feature selection methods have been proposed to improve the classification accuracy by reducing the original feature space. To improve the performance of text classification a new supervised feature selection approach has been proposed which develops a similarity between a term and a class. In this way every term will generate a score based on their similarity with all the classes and then all the terms will be ranked accordingly. The experimental results are presented on several TREC and Reuter data sets using knn classifier. The performances of the classifiers are compared using precision, recall, f-measure and classification accuracy. The proposed term selection approach is compared with document frequency thresholding, information gain, mutual information and chi(2) statistic. The empirical studies have shown that the proposed method performs significantly better than the other methods.