Amazon Sentiment Analysis

​

Project Type: Group ProjectTechnologies Used: Python, TensorFlow/Keras, NLP, LSTM, Bidirectional LSTM, GloVe Embeddings, Pandas, Matplotlib, Seaborn, SklearnDataset: Amazon Customer Reviews

Overview

​

This project involved building a sentiment analysis model to classify Amazon customer reviews into positive, neutral, or negative categories. The goal was to develop an accurate and efficient deep learning model using Bidirectional LSTMs and GloVe embeddings to enhance text understanding and sentiment classification.

​

Key Contributions

​

✅ Model Development: Implemented a Bidirectional LSTM model using TensorFlow/Keras for improved sentiment classification.✅ Data Preprocessing: Cleaned and tokenized text data, applied GloVe embeddings, and padded sequences for consistent input length.✅ Hyperparameter Tuning: Optimized LSTM layers, dropout rates, batch sizes, and optimizers to improve model performance.✅ Evaluation & Comparison: Achieved the highest accuracy (87.45%) among tested models, outperforming traditional machine learning approaches.✅ Visualization: Generated confusion matrices and classification reports to analyze model performance and identify misclassifications.

​

Results

​

📌 Final Model Accuracy: 87.45%📌 Validation Accuracy: 85.92%📌 Loss: 0.1824 (train), 0.2157 (validation)📌 Confusion Matrix: Demonstrated strong classification performance across all sentiment categories

​

🔗 Interactivity & Links

​

📌 GitHub Repository:(https://github.com/mqarmout/amazon_sentiment_analysis.git)

​