Sentiment-Based Product Recommendation System (Capstone)

🚀 Capstone project developed by a 4-member team.
🛠 Dockerization and container deployment independently implemented by Shibani Roychoudhury.

---

📚 Project Overview

• Goal: Build a recommendation system that suggests products to users based on their sentiment expressed in reviews.
• Approach:
  • Built a Sentiment Analysis Model using textual features from reviews.
  • Integrated User-User Collaborative Filtering for personalized recommendations.
  • Dockerized the full system for consistency across environments (solo contribution).

---

🛠 Key Highlights

• Used XGBoost Classifier with tuned hyperparameters for sentiment classification.
• Focused on data leakage prevention and meaningful feature engineering.
• The Collaborative Filtering component uses User-User similarity to recommend products based on inferred sentiment and user behavior.
• ✅ Dockerized and deployed the system using custom scripts.

---

🚀 Docker Deployment

• The full system is containerized and can be run with Docker.
• Docker image pushed to Docker Hub.
• Deployed briefly on Railway.app for public testing.

🔗 Docker Hub: helloshibani/sentiment-recommendation

🛠 Dockerization independently contributed by Shibani Roychoudhury.
📄 See docker_notes.md for build/run instructions.

---

🛠 Tech Stack

• Python
• Pandas, NumPy, XGBoost
• FastAPI + Jinja2
• Docker, Uvicorn
• Railway (optional deployment)

---

📈 How to Run

1. Clone the repository:

   git clone https://github.com/helloshibani/Sentiment-Based-Product-Recommendation-Analysis.git
   cd Sentiment-Based-Product-Recommendation-Analysis

2. Install dependencies:

   pip install -r requirements.txt

3. Build and run the Docker container:

   docker build -t sentiment-recommendation-system .
   docker run -p 8000:8000 sentiment-recommendation-system

Then open http://localhost:8000 in your browser.

---

📁 Folder Structure

• Sentiment_Recommendation_Capstone.ipynb: Main notebook
• app.py: FastAPI app logic
• model.py: Sentiment + recommendation engine
• templates/: Contains Jinja2 HTML files
• pickle/: Trained models (excluded in repo)
• Dockerfile, Procfile, requirements.txt, runtime.txt: For Docker & Railway deployment

---

📊 Results and Insights

• Predicts sentiment from user reviews and recommends relevant products.
• Containerized setup allows reproducible testing across platforms.

---

🚧 Future Work

• Improve model robustness and runtime performance.
• Streamline model loading in Docker containers.
• Link with Revised Solo Repo for enhanced model and improved explainability.

---

🙌 Acknowledgements

• Thanks to the Capstone team for collaborative effort and Railway deployment.
• Docker setup and deployment independently implemented by Shibani Roychoudhury

---

🔗 Connect

📬 LinkedIn – Shibani Roychoudhury
Let’s build, break, and learn together 🚀