Personalized preference elicitation in recommender systems using matrix factorization
Kirk Iserman, Yuhong Liu · 2017
Recommender systems are an effective way to find items of interest among the many available items for users based on their preferences. However, it is challenging to provide accurate recommendations for new users without any records in the system. We seek to provide a new method for building initial user profiles through personalized preference elicitations. In addition, to ensure limited amount of user effort involved in the preference elicitation process, matrix factorization approach is employed so that only a few representative latent factors are extracted to represent the massive amount of items available in the system. In comparison with another existing approach, our method significantly improves the accuracy of the initial user profile while requiring limited amount of user effort.