TensorFlow 1.x Deep Learning Cookbook

Getting ready

To make a recommender system using collaborative filtering, we need to modify our data. As an illustration, we will use the movie dataset taken from https://grouplens.org/datasets/movielens/. The data consists of two .dat files: movies.dat and ratings.dat. The movies.dat file contains three columns: MovieID, Title, and Genre for 3,883 movies. The ratings.dat file contains four columns: UserID, MovieID, Rating, and Time. We need to merge these two data files such that we are able to build an array, where, for each user, we have a rating for all the 3,883 movies. The problem is that users normally do not rate all movies, so we have non-zero (normalized) ratings only for some movies. The rest is made zero and hence will not contribute ...