MACHINE LEARNING-BASED SENTIMENT ANALYSIS OF PRODUCT REVIEWS

Abstract

In digital commerce, understanding consumer sentiment is crucial for businesses to make informed decisions and enhance customer satisfaction. With millions of reviews generated daily, consumers and manufacturers struggle to process this vast data. Hence, this research uses exploratory data analysis tools for the classification of product reviews into positive, negative, and neutral sentiments. Furthermore, three supervised machine learning techniques (Naïve Bayes, Logistic Regression, and Random Forest) are applied on Samsung Amazon reviews from Kaggle, conducting extensive training, testing, and evaluation. The research findings show that Random Forest outperformed the other models, achieving over 90% accuracy. This research offers an efficient solution for sentiment analysis, providing businesses with valuable insights to support product development and marketing strategies.