Information Recommendation Method Research Based on Trust Network and Collaborative Filtering

Abstract

Information recommender system is considered to be one of the most effective tools to solve the problem of information overload. Collaborative Filtering (CF), which utilizes similar neighbors to generate recommendations, is believed to be the most widely implemented and most mature technique for recommender systems. However, the recommendation results are often unsatisfactory due to the data sparsity of the input ratings matrix. Consequently, a hybrid recommender system which combines social network, trust network, and improved CF is proposed to enhance the accuracy of recommendation and overcome the weakness of data sparsity. Another advantage of the system is that utilizing the community structure discovered in social network as a new trust network sharply reduces the computation required for traditional CF. An empirical evaluation on Epinions.com dataset shows that the hybrid recommender system which incorporates social network and trust network into improved CF is more effective in terms of accuracy. This is especially evident on users who provided few ratings.