This project showcases a comprehensive approach to sentiment analysis of Amazon customer reviews, utilizing advanced NLP techniques and various machine learning models.
- Model Training with XGBoost & Logistic Regression: Implemented XGBoost and Logistic Regression models for sentiment classification.
- Text Processing & Labeling: Preprocessed review text using NLTK and labeled the sentiment with VADER sentiment analysis.
- BERT Model Training: Shifted to BERT for superior accuracy after initial attempts with traditional models. Fine-tuned a pre-trained BERT model for optimal performance.
- Deployment: Integrated a Flask server with React frontend for user interaction. The Flask server processes sentiment predictions, while React serves as the interface for real-time inputs.
- Traditional HTML & JS Setup: Developed a traditional HTML/JS setup for simpler engagement (Flask server-client communication).
- The trained BERT model file size is approximately 400MB, and it is not uploaded. Contact me if you would like access to the model.
- SMOTE was applied to address data imbalance during training, but BERT ultimately provided the most accurate predictions.
- Data Processing: Python (NLTK, VADER)
- Models: XGBoost, Logistic Regression, BERT
- Backend: Flask
- Frontend: React, Tailwind CSS
- Traditional Setup: HTML, JS
- Data Preprocessing: Cleaned the text data using NLTK and used VADER sentiment analysis to label sentiment.
- Model Training: Initially trained with XGBoost and Logistic Regression, but shifted to BERT for better accuracy.
- Deployment: Created a Flask API to handle sentiment prediction requests and integrated it with a React frontend for interactive use.
This project demonstrates end-to-end workflow from data preprocessing, model training (Logistic Regression, XGBoost, BERT), to deployment using Flask and React. Real-time sentiment prediction is now possible, giving users an interactive way to analyze customer reviews.