Transfer Learning to Extract Features for Personalized User Modeling

Aymen Ben Hassen, Sonia Ben Ticha


Personalized Recommender Systems help users to choose relevant resources and items from many choices, which is an important challenge that remains actuality today. In recent years, we have witnessed the success of deep learning in several research areas such as computer vision, natural language processing, and image processing. In this paper, we present a new approach exploiting the images describing items to build a new user’s personalized model. With this aim, we use deep learning to extract latent features describing images. Then we associate these features with user preferences to build the personalized model. This model was used in a Collaborative Filtering (CF) algorithm to make recommendations. We apply our approach to real data, the MoviesLens dataset, and we compare our results to other approaches based on collaborative filtering algorithms.


Paper Citation