Sentiment Analysis of Amazon Customer Reviews Using Machine Learning Models

Abstract

Sentiment analysis of Amazon customer reviews has become more important in today's digital marketplace, where understanding user mood directly impacts business strategies, product improvements, and customer satisfaction. Millions of reviews are created by everyday manual analysis, and there is an increasing demand for appropriate, automatic, and accurate ML solutions. This study addresses this need by implementing and comparing five ML models, which are D.T., which has 82.34% accuracy, Random Forest, which has 89.53% accuracy, Logistic regression, which has 91.52% accuracy, AdaBoost has 83.43% accuracy and XGboost, with 90.1% accuracy, to classify reviews as positive or negative. For the imbalanced dataset, the SMOTE technique was applied to balance sentiments. To address uneven distribution in mood analysis, SMOTE was used in this study. These discoveries provide businesses with actionable insights to automate review analysis, identify customer complaints, and make data-driven choices to boost products and services. This aims to classify user feedback into positive or negative categories. We trained our models on a dataset of 30,847 Amazon customer reviews covering various products and genres. This study shows the scalability of ML in actual-world mood categorization tasks and adds to the expanding body of work on applications. We also discuss the importance of striking a balance between computational effectiveness and model interpretability, particularly for parts that rely on illegal insights from massive amounts of unstructured feedback.