Bayesian Recommender Systems

With the proliferation of data nowadays, recommendation systems have become instrumental in filtering content for the user. These systems curate a set of personalized items to increase user satisfaction. In movie recommendation systems, the algorithm searches for content that would increase the user’s watch time. In this project, we design a Bayesian model to tackle the problem. Our goal is to predict the likelihood of a user liking an item. Experiments showed that our model is able to perform competitively with machine learning models. Moreover, in high confidence predictions, it surpasses them. However, the computational cost and lack of scalability of our model currently pose a limitation to its usage in a large-scale production setting.