A Comparative Performance Study of Machine Learning Algorithms for Sentiment Analysis of Movie Viewers Using Open Reviews

Abstract

The Internet facilitated the easy access to public opinions and reviews for any product or services. The collective opinions of people are significantly helpful for making decisions about any product or services. Movies are one of the most captivating pass times of the modern world on which people like to give their opinion/review. Movie reviews are personal opinions or comments shared via social media tool by a common viewer who has watched the movie. It provides an opportunity to know the outreach and response of the viewers to any film. Movie reviews influence the decision of prospective viewers as well as help producers, directors, and other stack holders for improving the quality aspect of the movie. Therefore, movie reviews play an important role in sentiment analysis of target viewers. In this research work, the sentiment analyses of viewers based on movie reviews using machine learning methods are discussed. The raw movie reviews are collected, and after performing preprocessing, features are extracted using bag of words, TF-IDF, bigram methods from text reviews. Various machine learning techniques including Naive Bayes classifier, Support Vector Machine, Decision trees, and ensemble learners are used with different feature extraction schemes to obtain a sentiment analysis model for positive or negative polarity in the movie review data sets. The performance of learners based sentiment analysis model is evaluated using accuracy, precision, recall, and f-measures. The objective of this research is to find the best classifier to test the reviews of movies given out by people so that we would know the overall general opinion of the audience. It is concluded that the set of classifiers can be used collaboratively to get effective results. Changes can be made from the very algorithmic level of the classifiers to gain better performance in the domain of study.