A Rating Prediction Method for Combining Social Network and Context Information

Recommendation accuracy can be improved by incorporating user relationships and context information. In this paper, we propose a rating prediction method incorporating social network information and context information to predict a rating value for unrated items. Firstly, we handle context information by Tensor Factorization to User-Item-Context tensor. Secondly, in order to incorporate social network information, we introduce an individual-based social regularization term to Matrix Factorization (MF) model to predict missing rating value of a user for an item by learning opinions from user's friends who have similar tastes. Lastly, we apply Linear Weighting Method to fuse predictive values based on social networks and context information. Experimental results on the real datasets show that in most cases, our approach gains better results than already widely used method.