Simple Content-based Recommendation Engine API With Flask [Heroku Deployed]

A simple content-based Netflix shows recommendation engine run as an API with Flask and deployed to Heroku. Medium article: https://medium.com/@MAbdElRaouf/simple-content-based-recommendation-engine-flask-api-heroku-dd27760dfe8e

Installation

1. Create new folder named "Netflix Shows Recommendation API"

2. Open a new command (or Anaconda) prompt inside the folder, or point terminal directory to its path:

   cd /d <parent directory path>\Netflix Shows Recommendation API

3. Create a new Python virtual environment:

   virtualenv recommendation_api_env

4. Activate recommendation_api_env:

   recommendation_api_env\Scripts\activate

5. The packages required for this project are pandas, scikit-learn, flask, gunicorn and flasgger. Run the following command to batch install them:

   pip install pandas sklearn flask gunicorn flasgger

6. After it has finished installing, save the project's list of packages to a text file with this command. Heroku uses this file as reference to what packages to install:

   pip freeze > requirements.txt

7. Install Heroku CLI: https://devcenter.heroku.com/articles/heroku-cli

8. Create Procfile for Heroku declaring gunicorn as the process type:

   echo web: gunicorn recommendation_api:app > Procfile

Usage

• Run shows_vectorizer.py to generate a pickle file of TF-IDF vectorizer.

• Start the recommendation engine API locally by running recommendation_api.py.

• To run it on Heroku's cloud, see the setup instructions here.

• Communicating with the API can be either through sending JSON POST requests to endpoint /api/ or interacting with the Swagger UI at endpoint /apidocs/

  • Example 1: POST request to https://netflix-recommendation-api.herokuapp.com/api with payload:

    {
    "title": "The Matrix"
    }
    

    is responded to with

    {
     "result": [
       {
         "title": "The Matrix Reloaded",
         "confidence": 0.4
       },
       {
         "title": "The Matrix Revolutions",
         "confidence": 0.3
       },
       {
         "title": "Jupiter Ascending",
         "confidence": 0.1
       },
       {
         "title": "Terminator 3: Rise of the Machines",
         "confidence": 0.1
       },
       {
         "title": "Sense8",
         "confidence": 0.1
       }
     ]
    }
    

  • Example 2: Using the user interface at https://netflix-recommendation-api.herokuapp.com/apidocs

Built With

• Scikit-Learn
• Flask
• Flasgger