Personalised Product Recommendation System

This repository contains a personalised product recommendation system that leverages FAISS for similarity-based recommendations and FastAPI for deployment. The application provides recommendations for users based on content-based and collaborative filtering approaches and ensures diversity and relevance in suggestions.

Features

• Content-Based Filtering: Uses product features like ratings and discounts to recommend similar items.
• Collaborative Filtering: Recommends products by analysing user-product interactions.
• Hybrid Recommendation System: Combines content-based and collaborative filtering for better accuracy and diversity.
• FAISS Integration: Utilises FAISS for efficient similarity search.
• REST API: Built with FastAPI to provide endpoints for querying recommendations.
• Dockerised Deployment: Includes Docker support for easy deployment.

Installation

1. Clone the repository:

   git clone https://github.com/Aubur9y/Personalised-Product-Recommendation-System.git
   
   cd Personalised-Product-Recommendation-System
   

2. Build and run the Docker container:

   docker build -t personalised-recommendation-app .
   
   docker run -d -p 8000:8000 personalised-recommendation-app
   

3. Alternatively, set up the environment locally:

   python3 -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   pip install -r requirements.txt
   uvicorn app:app --reload
   

API Endpoints

Get Recommendations:

• URL: /recommend/
• Method: GET
• Parameters:
  • user_index (int): The index of the user to get recommendations for.
  • top_n (int): Number of recommendations to return.
• Example:

  curl "http://localhost:8000/recommend/?user_index=0&top_n=5"

• Response

  {
      "user_index": 0,
      "recommendations": [
          {"product_id": 132, "score": 1.0},
          {"product_id": 113, "score": 0.9}
      ]
  }
  

Future Enhancements

• Integrate Apache Spark for scalable collaborative filtering.
• Support for dynamic user interaction and feedback integration.
• Improved diversity logic for better recommendations across different product categories.