Enhancing Amazon Online Product Review Sentiment Analysis by LSTM-Based Deep Learning Framework

Consumers are increasingly relying on online customer reviews for assistance in their purchasing choices, rendering sentiment analysis a critical tool for understanding feedback. E-commerce platforms like Amazon enable users to share insights that influence buying behaviors and provide businesses with valuable data on consumer preferences. For extracting meaningful insightful data from large volumes of reviews, classifying sentiments into positive and negative categories is essential. Sentiment analysis, a pivotal task in Natural Language Processing (NLP), helps businesses enhance customer experiences and support data-driven decisions. This study analyzes customer reviews from the amazon platform using a combination of machine learning (ML) and deep learning approaches (DL). Extensive preprocessing was conducted, including a VADER lexicon-based sentiment analyzer from NLTK and the TF-IDF method for feature extraction to optimize model performance. We compared four ML algorithms—Logistic Regression (LR), Decision Tree (DT), Multinomial Naive Bayes (MNB), and XGBClassifier—with a DL-based LSTM model. Our results demonstrate that the LSTM model outperformed all other approaches, yielding an accuracy of 98.23%, precision of 98.20%, recall of 98.16%, and an F1-score of 98.16%. These outcomes underscore the efficiency of combining advanced preprocessing techniques with DL models for sentiment analysis in e-commerce.