Incorporating collaborative tagging in social recommender systems

Abstract

Recommender systems play a major role in recent times to help online users in finding the relevant items. Traditional recommender systems have been analysed immensely, but they ignore information like social friendships, tags which when incorporated in recommendation can improve its accuracy. With the advent of social networking sites, study of social recommender systems has become active. Most of the users ask their friends for recommendation. But not all friends have similar taste as that of the user, and different group of friends contribute to different recommendation tasks. So this paper proposes an approach to identify different group of friends by grouping users based on items and retains the personal interest of experienced by incorporating individual-based regularization in basic matrix factorization. Information like ratings, tags and friendship are used in predicting the missing values of user-item matrix efficiently. Empirical analysis on the dataset proves that the proposed approach is better than the existing methods.