Sentiment Analysis of Amazon Customer Review by Using Machine Learning Models

Sentiment analysis is a powerful tool for understanding customer opinions and enhancing business decision-making processes. This study explores the sentiment analysis of a large dataset consisting of half a million Amazon customer reviews, categorizing them into three sentiment classes: positive, negative, and neutral. The research employed five machine learning models -Random Forest Classifier (RFC), Logistic Regression Classifier (LRC), Bernoulli Naive Bayes (BNB), Linear Support Vector Classifier (L-SVC), and Multinomial Naive Bayes (MNB)trained using two feature extraction techniques, Bag of Words (BoW) and Term Frequency Inverse Document Frequency (TFIDF), to evaluate their effectiveness in accurately classifying sentiments. The models were evaluated using standard metrics and a time-adjusted efficiency formula balancing accuracy and computational cost. RFC achieved the highest accuracy despite higher time cost, while L-SVC offered a strong balance of accuracy and efficiency. A detailed discussion highlights its advantages and limitations, providing valuable insights into its practical applications. To extend the practical value of this research, a userfriendly application for real-time sentiment analysis of customer reviews was developed, showcasing the seamless integration of machine learning models into practical use. It highlights the importance of selecting suitable models and features for accurate, efficient sentiment classification, contributing to research and practical advancements in large-scale sentiment analysis.