Polarity classification on web-based reviews using Support Vector Machine
Renato S. C. da Rocha, Leonardo Forero, Harold de Mello, Manoela Kohler, Marley M. B. R. Vellasco · 2016
This paper presents a hybrid filter-wrapper approach to sentiment polarity classification. It is a two-phase feature selection method that differs from other approaches in a main aspect: an initial preprocessing step with typical data mining filters. After applying such filters, a wrapper-based feature selection method, which integrates the genetic algorithm and the support vector machine classifier, is used for sentiment analysis on the Internet Movie Database (IMDb) reviews. We show that this approach improves the classification accuracy compared to other methods, including those such preprocessing techniques are applied separately to the database.