Clustering based feature selection using Extreme Learning Machines for text classification

摘要: The expansion of the dynamic Web increases digital documents, which has attracted many researchers to work in field text classification. It is an important and well studied area machine learning with a variety modern applications. A good feature selection paramount importance increase efficiency classifiers working on data. Choosing most relevant features out what can be incredibly large set data, particularly for accurate This paper motivation that direction where we propose new clustering based technique reduces size. Traditional k-means along TF-IDF Wordnet helps us form quality reduced vector train Extreme Learning Machine (ELM) Multi-layer ELM (ML-ELM) have been used as experimental carried 20-Newsgroups DMOZ datasets. Results these two standard datasets demonstrate our approach using ML-ELM over state-of-the-art classifiers.