This experiment utilizes collaborative and content-based filtering and matrix factorization to create a movie recommender engine.
Using the Matchbox Recommeder module, I have trained, scored, and evaluated 4 different metrics to create a list of movie recommendations for each user. To evaluate the accuracy of the predictions the Evaluate Recommender module compares the predictions of a recommendation model with the corresponding “ground truth” data by computing the average normalized discounted cumulative gain (NDCG) for each model. In this experiment content-based filtering based on related items performed best. The training data is approximately 225,000 ratings for 15,742 movies by 26,770 users, extracted from Twitter using techniques described in the original paper by Dooms, De Pessemier and Martens. The paper and data can be found on GitHub.