🌱 EcoPack-AI

AI-Based Sustainable Packaging Recommendation System built with Flask, Machine Learning, and PostgreSQL.

🚀 Project Overview

EcoPack-AI recommends the most suitable packaging material based on:

• Product category
• Fragility level
• Shipping type
• Sustainability priority

The system predicts and ranks materials using:

• Packaging cost
• CO₂ impact
• Material suitability score

Recommendations are stored in the database and visualized in a BI-style dashboard with analytics and export options.

🧠 Core Features

• Intelligent recommendation engine with category + fragility filtering
• Dynamic weighted scoring logic based on sustainability priority and shipping type
• Interactive dashboard analytics (distribution, ranking, trends)
• Recommendation history tracking with clear-history endpoint
• Export support (PDF / Excel on frontend)
• Environment-based DB configuration (PostgreSQL + SQLite fallback)

🏛 Architecture

The platform follows a layered architecture from UI → Flask API → ML models → PostgreSQL → BI dashboard.

🏗 Tech Stack

• Backend: Python, Flask, Flask-CORS
• Data/ML: Pandas, NumPy, Scikit-learn, XGBoost, Joblib
• Database: PostgreSQL (production), SQLite (local fallback), SQLAlchemy
• Frontend: HTML, CSS, JavaScript, Chart.js, Plotly
• Deployment: Gunicorn, Render

📁 Project Structure

EcoPackAI/
├── backend/
│   ├── app.py
│   ├── requirements.txt
│   ├── rf_cost_model.pkl
│   ├── xgb_co2_model.pkl
│   └── scaler.pkl
├── frontend/
│   ├── index.html
│   ├── style.css
│   ├── script.js
│   └── config.js
├── images/
│   ├── architecture_new.png
│   ├── dashboard-main.png
│   ├── material-distribution.png
│   ├── recommendation-trends.png
│   ├── model-metrics-cost.png
│   ├── model-metrics-co2.png
│   └── best-material-output.png
├── model_training.ipynb
├── CHANGELOG.md
├── CONTRIBUTING.md
├── LICENSE
├── Procfile
├── runtime.txt
└── README.md

🔌 API Endpoints

• GET / → Health check
• POST /recommend → Generate top material recommendations
• GET /history → Retrieve recommendation history
• POST /history/clear → Clear stored history
• GET /analytics → Dashboard analytics aggregates

⚙️ Environment Variables

Create a .env file using .env.example:

API_KEY=change_me
DATABASE_URL=postgresql://username:password@localhost:5432/ecopackai

Note: Current code requires DATABASE_URL for PostgreSQL; if not provided, it automatically uses local SQLite (ecopackai.db) for development.

💻 Run Locally

1. Open terminal in project root:

   cd backend

2. Create and activate virtual environment (Windows):

   python -m venv venv
   venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Start backend:

   python app.py

5. Open frontend file frontend/index.html in browser (or host via a simple static server).

☁️ Deployment (Render)

• Build Command

  pip install -r backend/requirements.txt

• Start Command

  gunicorn backend.app:app

• Required Environment Variables

  • DATABASE_URL
  • API_KEY (reserved for secured endpoint enhancements)

📊 Evaluation Summary

Models are evaluated using:

• MAE
• RMSE
• R² Score

Two model roles:

• Random Forest → packaging cost prediction
• XGBoost → CO₂ prediction used in ranking

Model Results Snapshot

Random Forest (Cost Prediction):

XGBoost (CO₂ Prediction):

🖥 Dashboard Screenshots

Main BI Dashboard

Material Distribution & Comparison

Sustainability Ranking & Recommendation Trends

Best Material Output (Notebook Validation)

🧩 Screenshot File Checklist

Place these exact files inside the images/ folder (same spelling and case):

• architecture_new.png
• dashboard-main.png
• material-distribution.png
• recommendation-trends.png
• model-metrics-cost.png
• model-metrics-co2.png
• best-material-output.png

👨‍💻 Author

Rajan Kumar

• GitHub: https://github.com/RajanKumar44

---

If you like this project, give it a ⭐ on GitHub.