Abstract
Recommender system is an information filtering system that provides suggestions for users with top N corresponding items based on ratings. Matrix Factorization is the basic idea to predict a personalized ranking over a set of items for an individual user with the similarities among users and items. Matrix Factorization is a widely used collaborative filtering method in recommender systems. In this work, a novel matrix factorization model with neural network architecture is proposed. Firstly, data is stored in a user-movie matrix with explicit ratings. With this matrix as the input, an embedding layer is used to create latent vectors that represent the features of both users and movies. Secondly, a deep structure learning architecture is applied to learn a common low dimensional space for the representations of users and items. Finally, a loss function based on mean square error (mse) is used and the goal is to minimize this loss between the predicted and target outputs. Model performance evaluation in different experimental settings using two datasets: Netflix Prize and MovieLens. The deep learning model can get 0,8051 RMSE (Netflix Prize) or 0,7565 RMSE (MovieLens).