Movie Recommender System Using Matrix Factorization and Deep Learning Techniques

Abstract

Recommendations drive so many of our decisions on a daily basis. Recommender systems help consumers to find new information, products and services tailored to their requirements. Recommendation engines use the feedback of users to find new relevant items for them or for others with the assumption that users who have made homogeneous choices in past are highly anticipated to make similar choices in forthcoming future. There are various types of recommendation systems like – non personalized recommender system, collaborative filtering based recommender systems, deep learning based recommender systems. The goal of our project is to predict the ratings that users may give to movies that they have not rated yet and to build and test various recommender systems and then finally minimize the root-mean-square-error between the projected user ratings and true ratings of the user using matrix factorization and Deep Learning techniques. The whole project is based on user movie ratings data, so we need to collect that data. We have collected the data from movielens website and then we have filtered and processed our data. The data is then split into training and test sets and the test set. To make our Matrix Factorization model, we have used SVD, SVD++ and related algorithms and also keras. For our Deep Learning models, we have built Autoencoders, Matrix Factorization and Residual Learning in keras. Finally, have calculated RMSEs for each of our recommender systems and compared them.