A Hybrid Recommender System with Deep Learning

Since the amount of information and products in virtual environments increased, the need for recommender system as a solution to help users and customers find and explore their needs become more important. Many companies such as Netflix and YouTube have been receiving huge amount of items and users at the same time and the need to satisfy users with good and accurate recommendations has forced them to design new and better recommender systems which can match user preferences better. Among different recommendation techniques, collaborative filtering (CF) is more popular and many works have done in order to decrease the challenges in this field. In other way Content based (CB) recommender systems are useful to model the item content and offers items to users based on the profile of user preferences about items and their contents. On the other hand deep learning approaches have made significant success in speech recognition, computer vision and also natural language processing. All thought in recent years there are good attempts about using the power of deep learning methods in both collaborative filtering and content based recommender systems, still there are few works modeling hybrid methods in deep learning. In this paper we design a hybrid recommender system that uses both collaborative filtering information (user-item rating matrix) and content information of items. We use MovieLens latest dataset for our model and since it has just collaborative information data, we gather other content information of movies from IMDB and use them as content data for our hybrid model. The evaluation results show that our proposed model outperforms state of the art models in rating prediction.