Movie Recommendation System using Auto-encoders

Abstract

This undertaking report gives a model for the rating expectation task in movie recommendation system which gives best predictions of ratings of users who have not give predictions in ratings dataset for any given movie.Our model is based on Collaborative Filtering technique which is based on past behaviour of user not the content. Our model depends on stacked auto encoder with 4 layers with arrangement 20-10-10-20 neurons and is prepared end-to-end with no layer-wise pre-training. We additionally decreased our test loss however much as could reasonably be expected via preparing model on 400 epochs. We have used MovieLens Dataset, which is most common dataset available on internet for recommendation purpose. The dataset contains(1M) 1,00,209 anonymous ratings.