An efficient technique for hybrid classification and feature extraction using normalization

Abstract

Text classification is technique for assigning the class or label to a particular document within predefined class labels. Predefined classes examples are sports, business, technical, education and science etc. Classification is supervised learning technique i.e. these classes are trained with certain features and then document is classified based on similarity measure with these trained document set. Text classification is used in many applications like assigning the label to the documents, separating the spam messages from the genuine one, filtering of text, natural language processing etc. Feature selection, extraction and classification are various phases for assigning label to any document. In this paper, PCA is used for feature extraction, ABC is used for feature selection and SVM is used for classification. PCA is improved by applying normalization-using size of features in our proposed approach. It reduces the redundant features to larger extent. There are very few research works, which have implemented PCA, ABC and SVM for complete classification. Evaluation parameters like accuracy, F-measure and G-mean are calculated to check classifier efficiency. The proposed system is deployed on 20-Newsgroup dataset. Experiment analysis proves that accuracy is improved using our proposed approach as compared to existing approaches.