Voting Ensemble Classifier for Sentiment Analysis

Abstract

This paper proposes use of ensemble voting learning classification approach for sentiment classification of movie review dataset collected from IMDB. Sentiment analysis is a technique used to extract opinions from text including reviews, social messaging, etc. Generally, the opinion is classified into positive and negative polarity. The approach has been used in variety of domains including financial, educational, and other areas. Over the years many researchers have worked on sentiment classification to predict the opinion of text using several machine learning algorithms. The work carried out in this paper proposes the use of ensemble voting learning approach in which Naïve Bayes (NB), K-nearest neighbor (KNN), random forest (RF), and decision tree (DT). The results are compared with individual classification algorithm and voting approach. We have also used normalized and denormalized data to compare the accuracy results. The results show that using voting approach decreases root mean square error and increases precision of the classifier. Maximum accuracy of 80.13% is obtained with voting and normalized data.