The objective of this project is to build a recommender system on user ratings data based on matrix factorization. The data contains a total of 150 jokes with 50692 users who have rated many of the jokes. The recommendation model was implemented using Graphlab in Python 2.7. A test rmse of 2.73 and a training rmse of 3.02 was obtained. Latent features or topics were also discovered in the jokes.
EDA on jokes and user ratings.ipynb: contains results from the exploratory analysis on the data, to gain more insights about the jokes and user ratings
recommender_system.py: uses Graphlab to create a matrix factorization based recommendation system