A New Approach for Selecting Informative Features For Text Classification

Selecting useful and informative features to classify text is not only important to decrease the size of the feature space, but as well for the overall performance and precision of machine learning. In this study we propose a new feature selection method called Informative Feature Selector (IFS). Different machine learning algorithms and datasets have been utilised to examine the effectiveness of IFS, and it is compared to well-established methods, namely Information Gain, Odd Ratio, Chi Square, Mutual Information and Class Discriminative Measure. Our experiments show that IFS is able to outperform aforementioned methods and to produce effective and efficient results.