Prodo - Advanced AI Product Recommendation System

A sophisticated product recommendation system that combines multiple AI approaches for personalized recommendations, including real Neural Collaborative Filtering (NCF) training.

---

Features

• Neural Collaborative Filtering (NCF): Deep learning-based collaborative filtering using PyTorch, trained on your user interactions.
• Content-based Recommendations: Uses BERT for text and CNN for image features.
• Graph-based Recommendations: Knowledge graph with user-product interactions.
• Time-aware Recommendations: Considers temporal patterns in user behavior.
• Hybrid Recommendation Approach: Combines all the above for robust, explainable recommendations.
• REST API: Easily integrate with any frontend or tool (e.g., Postman).
• Robust Data Handling: Works with or without timestamps in your interaction data.

---

Project Structure

prodo/
├── advanced_models.py         # Core AI models (NCF, content, graph, time-aware, hybrid)
├── app.py                    # Flask API endpoints
├── requirements.txt          # Project dependencies
├── data/                     # Data directory
│   ├── products.csv          # Product information
│   ├── users.csv             # User information
│   └── user_interactions.csv # User interaction data
├── models/                   # Directory for trained models (optional)
├── utils.py                  # Utility functions
├── test_utils.py             # Test scripts
├── test_recommendation_model.py
├── test_data_preprocessing.py
├── data_preprocessing.py
└── README.md

---

Setup

1. Clone the repository:

   git clone <your-repo-url>
   cd prodo

2. Install dependencies:

   pip install -r requirements.txt

3. Download spaCy model:

   python -m spacy download en_core_web_sm

4. (Optional) Download BERT weights and ResNet weights:
   These will be automatically downloaded the first time you run the app.

5. Prepare your data:
   Place your CSV files in the data/ directory. See below for format.

6. Run the application:

   python app.py

   The API will be available at http://127.0.0.1:5000/

---

Data Format

products.csv

product_id	name	description	category	price	image_url
101	Smart Watch Pro	Advanced fitness tracking smartwatch	Electronics	199.99	https://example.com/watch.jpg

users.csv

user_id	name	age	preferences	location
1	John Doe	28	Electronics,Gaming	New York

user_interactions.csv

user_id	product_id	clicks	rating	(timestamp, optional)
1	101	5	4	1718246400

• timestamp is optional. If present, enables time-aware recommendations.

---

API Endpoints

1. POST /recommend

Get personalized recommendations for a user.

• Request:

  • Headers: Content-Type: application/json
  • Body:

    {
      "user_id": 1,
      "context": {},
      "top_n": 5
    }

    • user_id: The user to recommend for.
    • context: (optional) Dictionary with context (e.g., time, location).
    • top_n: (optional) Number of recommendations (default: 5).

• Response:

  {
    "user_id": 1,
    "recommendations": [
      {
        "product_id": 101,
        "score": 0.92,
        "explanation": "Recommended based on your preferences and behavior patterns (confidence: 0.92)"
      }
    ],
    "context_used": { ... }
  }

How to use with Postman:

1. Set method to POST and URL to http://127.0.0.1:5000/recommend
2. In the Headers tab, add:
   • Key: Content-Type
   • Value: application/json
3. In the Body tab, select raw and choose JSON from the dropdown.
4. Paste the example JSON above and click Send.

---

2. POST /train

Retrain all models (including NCF) on the latest data.

• Request: No body required.
• Response:

  { "message": "Models retrained successfully" }

---

Model Training

• The NCF model is trained automatically on startup and whenever you POST to /train.
• Training progress and loss will be shown in the terminal.
• The system is robust to missing columns (e.g., timestamp).

---

Notes

• For best results, use real user interaction data with ratings.
• The project is for educational/demo purposes. For production, use a WSGI server and secure your endpoints.
• You can extend the system with more advanced models or data as needed.

---

License

MIT (or your chosen license)

---

Contact

For questions or contributions, open an issue or pull request on GitHub.