Content-Based Music Recommendation Using Non-Stationary Bayesian Reinforcement Learning

Literature Survey

The currently available literature on music recommendation systems can be broadly classified into three categories (X. Wang et. al., 2014):

• •

  Collaborative Filtering

• •

  Content-based Filtering

• •

  Context-based Filtering

Collaborative filtering (CF) methods are widely used in recommender systems. They provide recommendations based on ratings that users give to items. The results of these techniques are quite good; however, the difficulty in obtaining explicit feedback in the form of ratings from the users causes the sparsity problem, which takes place when the number of available ratings for the items to be recommended is small. CF is powerless when confronted with the new-song problem — it cannot recommend songs without prior usage history. (D. S. Moreno et. al., 2016, Y. Koren et. al., 2009).