K-means and Wordnet Based Feature Selection Combined with Extreme Learning Machines for Text Classification

Abstract

The incredible increase of online documents in digital form on the Web, has renewed the interest in text classification. The aim of text classification is to classify text documents into a set of pre-defined categories. But the poor quality of features selection, extremely high dimensional feature space and complexity of natural languages become the roadblock for this classification process. To address these issues, here we propose a k-means clustering based feature selection for text classification. Bi-Normal Separation (BNS) combine with Wordnet and cosine-similarity helps to form a quality and reduce feature vector to train the Extreme Learning Machine (ELM) and Multi-layer Extreme Learning Machine (ML-ELM) classifiers. For experimental purpose, 20-Newsgroups and DMOZ datasets have been used. The empirical results on these two benchmark datasets demonstrate the applicability, efficiency and effectiveness of our approach using ELM and ML-ELM as the classifiers over state-of-the-art classifiers.