In my Bachelor thesis I examine the world of matrix reconstruction applied to the recommender system problem. The methods for reconstruction I examine are Nuclear Norm Minimization, Restricted Boltzmann Machine, and k-Nearest Neighbor. I apply these to synthetic non complete low rank matrices and datasets obtained from movielens. I also look at a possible implementation of a location based recommender system in a store, which engages customers by suggesting items to buy based on their physical location in the store.
Click the link below to read the .pdf version.