Personalized Context-Aware Rating Prediction Model and Recommendation Approach Based on Neural Network

Recent years, context has been identified as an important factor in recommender systems. Great contributions have been done for context-aware collaborative filtering recommendation approaches, but the contextual parameters in current approaches have same weights for all users. In the paper a recommendation approach based on BP neural network is proposed to learn a personal context-aware rating prediction model for each user. Each input unit in the model represents a contextual parameter and the weights of neural units are different for every user. Finally, we evaluate experimentally our approach and compare it to context-based collaborative filtering and Slope One. The experimental results show our algorithm out performs Slope One and traditional context-aware collaborative filtering.