Sentiment Analysis of Amazon Customer Review by Using Deep Learning Models

The abundance of user-generated text in the digital age provides valuable insights into opinions, preferences, and sentiments. This study presents a comprehensive sentiment analysis system for Amazon customer reviews using deep learning models, including Long Short-Term Memory (LSTM), Bidirectional LSTM, Recurrent Neural Networks (RNN), Gated Recurrent Units (GRU), Attention, and FastText. The proposed system is designed with an interactive customer interface that allows user to enter custom reviews to receive sentiment classifications: positive, negative, or neutral. The system evaluates model performance using key metrics such as precision, recall, F1-score, accuracy, validation accuracy, loss, and validation loss. Extensive experimentation reveals that the Attention model outperforms traditional architectures by capturing long-range dependencies and context in review text, while FastText demonstrates superior speed in processing large datasets. The results indicate that model selection and architecture tuning play a critical role in optimizing sentiment analysis performance. The developed system not only automates sentiment assessment but also provides robust insights into customer opinions, aiding businesses in decision-making processes and enhancing customer experiences. This research highlights the potential of deep learning models in transforming unstructured review data into actionable insights and contributes a user-centric tool for real-time sentiment analysis.