Motivation: Recommendation systems are used by major services like YouTube, Netflix and Amazon to recommend items or content of interest to their users. Sequential patterns in data play a significant role in how well a recommendation system performs. However, the struggle to uncover complex sequential relationships in a user’s history is still common.
Results: In this study, we propose a deep learning based approach that utilizes the user’s history. Using a multiplicative long short term memory (mLSTM), we capture the sequential information of a user. On the MovieLens 1M dataset of 6040 users and 3706 movies, resulting in over a million interactions, we trained a deep learning network to capture the sequential information of each user by utilizing the user characteristics, the item features and the user-item history. Our highest reported Mean Squared Error (MSE) was 0.958.