3D visualization of sentiment measures and sentiment classification using combined classifier for customer product reviews

Abstract

The Internet has wide reachability making many users to buy the products online using e-commerce websites. Usually, users provide their opinions, comments, and reviews about the products in social media, e-commerce websites, blogs, etc. The product review comments provided by the customers have rich information about the usage of the products they bought and their sentiments towards those products. In this research, we have collected reviews from Amazon.com and performed sentiment analysis to collect sentiment information. We have proposed 3D visualizations to represent sentiment information, such as sentiment scores and statistics about words used in the reviews. The 3D visualizations are useful to represent large sentiment related information and to have an in-depth understanding of sentiments of users. We have developed a combined classifier using Logistic Regression, Decision Tree and Support Vector Machine. From the reviews, we formed N-gram features using a bag of words and performed sentiment classification using combined classifier. On 10 fold cross-validation, a maximum classification rate for combined classifier of 90.22% is obtained for sentiment classification.