Recommendation System

This project presents an end-to-end recommendation system designed for an e-commerce platform. The system utilizes item-based collaborative filtering using the Surprise library for recommendation generation and Flask for creating APIs for easy integration into Android apps or websites.

Implementation Details

1. Main Driver - myapp.py

• This is the core of the project, serving as the main driver file.
• It configures and deploys the Flask API, making it ready for use on IIS servers.

2. Similarity Matrix Calculation - InsertRecSys.py

• This program is responsible for calculating the similarity matrix based on user-item interactions.
• It then inserts this matrix into a MongoDB database for later use in recommendation generation.

3. Recommendation Generation - RecSys.py and RecSys2.py

• RecSys.py:

  • These programs generate top-N recommendations by reading and utilizing the similarity matrix stored in the MongoDB database.
  • Users can specify the item ID and the number of recommendations they desire as parameters in the API.

• RecSys2.py:

  • Similar to RecSys.py, these programs generate top-N recommendations from the MongoDB database using the similarity matrix.
  • In addition to item ID and the number of recommendations, users can also provide a warehouse ID (wid) parameter to restrict recommendations to items available in their city's warehouse.

Evaluation

• The recommender system has been rigorously evaluated using Leave-One-Out (LOO) Cross Validation.
• Achieved a hit rate of 21%, which is considered quite good in the context of recommendation systems.

Usage

To leverage the recommendation system and its APIs, follow these steps:

1. Ensure you have the necessary dependencies installed, including Flask, Surprise, and MongoDB drivers.
2. Run myapp.py to configure and deploy the Flask API.
3. Calculate and insert the similarity matrix into MongoDB using InsertRecSys.py.
4. Utilize the recommendation APIs (RecSys.py and RecSys2.py) in your Android app or website by passing the appropriate parameters for item ID, number of recommendations, and, if needed, the warehouse ID.

Acknowledgments

This project showcases the power of item-based collaborative filtering and Flask APIs for enhancing user experience on your e-commerce platform.