Recommendation algorithm based on users’ interest preferences and Restricted Boltzmann machine

Abstract

In order to overcome the disadvantages of recommendation algorithm based on restricted Boltzmann machine, which only uses the severely sparse scoring matrix in the training process, and can not effectively extract user characteristics effectively, and also ignores the users' interests and item attributes. a recommendation algorithm based on users' interest preferences and Restricted Boltzmann Machines is proposed. Firstly, the Restricted Boltzmann machines for collaborative filtering algorithm is utilized to predict the user rating of the item is r ₁ ; then the information of users' interest preference is used to establish a users' rating preference model, and predict the user rating the item is r ₂ ; Finally, the linear regression algorithm is utilized to confirm the weights of r ₁ and r ₂ to obtain the final prediction score r . The experimental results show that the proposed algorithm in this paper can improve the predicted accuracy of the recommendation system.