Poisson Matrix Factorization For TV Recommendations

Recommendation systems are becoming more and more popular within e-commerce websites to help drive user engagement. It is not just limited to e-commerce though, websites such as Netflix or Spotify utilize recommendation systems to better engage users in movies and TV shows, or music. This thesis explores the mathematics and assumptions behind recommendation systems, such as how data is distributed and different algorithms used. The thesis then performs a case study on Reddit TV show data to build a recommendation system. To improve the results of the recommendation system, this thesis makes changes to a Python Recommendation System Library to enable Poisson Factorization. The changes proposed can be integrated into the existing Python library, helping other programmers make more meaningful and accurate recommendations.