Hybrid Recommendation System

This project is a hybrid recommendation system built with Python, Fast API, scikit-learn, pandas, numpy, and more. It uses a combination of collaborative filtering, content-based filtering, and popularity-based filtering to generate recommendations.

Features

• Hybrid recommendation system with adjustable weights for each component
• Supports cold start recommendation for new users
• Can be configured with environment.yml (conda), requirements.txt (plain python), or Dockerfile (docker)

Usage

With Conda

1. Install conda if you haven't already
2. Run conda env create -f environment.yml to create a new environment
3. Activate the environment with conda activate recomender
4. Run python main.py to start the API

With plain Python

1. Install the requirements with pip install -r requirements.txt
2. Run python main.py to start the API

With Docker

1. Build the docker image with docker build -t recomender .
2. Run the container with docker run -p 8000:8000 recomender
3. The API will be available at http://localhost:8000

API Endpoints

GET /users/{user_id}/recommendations

Returns a list of recommended products for the given user

GET /products/{product_id}/similar

Returns a list of similar products to the given product

GET /products/popular

Returns a list of popular products

GET /products/new

Returns a list of new products

POST /users/{user_id}/interactions

Creates a new interaction for the given user and product

Configuration

The following environment variables can be used to configure the API:

• RECOMENDER_CF_WEIGHT: The weight of the collaborative filtering component (default: 0.5)
• RECOMENDER_CB_WEIGHT: The weight of the content-based filtering component (default: 0.3)
• RECOMENDER_PB_WEIGHT: The weight of the popularity-based filtering component (default: 0.2)
• RECOMENDER_COLD_START_THRESHOLD: The minimum number of interactions required for a user to be considered "warm" (default: 10)
• RECOMENDER_COLD_START_RECOMMENDATIONS: The number of recommendations to return for cold start users (default: 10)

Development

To develop the API, run python main.py and use a tool like curl or a web browser to test the endpoints.