A fuzzy aproach for determination trust threshold in recommender systems based on social network

Trust-aware recommender systems are intelligent technology applications that make use of trust information and user personal data in social networks to provide personalized recommendations. Trust clearly elevates accuracy of prediction rating for users; especially, when the user's preferences are insufficient, trustworthy friend or people viewpoints can be confidence referenced. However in previous researches, Trust network is reconstructed by removing trust links between users having correlation coefficient below a specified threshold value. On the other hand it does not consider the ratings from friends below a certain trust value. It reduces usefulness data points in recommendation process and therefore raises the accuracy. So determination of the threshold value constitutes one of the main challenges with these approaches. As the performance of proposed methods are highly sensitive to this threshold. In this research, we propose a novel method for recommendation based on fuzzy logic in social networks. In this method the weighted average of all ratings are calculated in the process of recommendation. These weights are determined through a fuzzy method. Finally Epinions dataset is used to demonstrate the performance of proposed method.