Recommendr

Movie recommendation engine.

• Free software: BSD license
• Documentation: http://recommendr.rtfd.org.

Features

• A Redis backend for storing movie and rating data.
• A simple user-based recommendations algorithm with swappable distance functions.
• Item-based recommendation algorithm in work.
• A demo command-line client.

Try it Out

• Clone the repo:

  git clone git@github.com:chrislawlor/recommendr.git
  

• Create a python virtual environment with virtualenvwrapper:

  mkvirtualenv recommendr
  

• Install requirements:

  pip install -r requirements.txt
  

• Install recommendr:

  python setup.py install
  

• First, import some MovieLens data into Redis:

  python data/import_data.py
  

• Run the demo program:

  python demo.py
  

The demo program will ask you for ratings until you have rated 5 movies, then it will give some recommendations. Recommendations should improve the more times you run the demo program.

NOTE: If your Redis instance is somewhere other than locahost:6379, set the REDIS_HOST and REDIS_PORT environment variables. If you wish to use a Redis DB other than 1, set REDIS_DB.

Key Code Points

recommendr.db: Implements a Redis DB backend suitable for storing movie and rating information

recommendr.get_user_based_recommendations(reviewer_id, num=20, similarity=sim_distance)

returns the top recommendations for a given user. It defaults to using Euclidean distance for the similiarity function, optionally pass recommendr.similarity.sim_pearson to use the Pearson Coefficient.

Test Suite

I haz one:

python setup.py test