Leveraging Global Information In Social Network For Recommendation Algorithm

Nowadays, the Internet severely suffers from the information overload problem due to the explosive growth of digital information, which greatly hinders the customers in obtaining their interested information efficiently. Recommender systems hence becomes more and more prevalent and important for customers. However, there are still several key problems to be solved, such as how to improve the accuracy and address data sparsity and cold start problems of recommendation. To tackle these issues, this paper proposes a new collaborative filtering recommendation algorithm based on global information, namely the whole user-item rating matrix, which is improved by traditional similarity methods to further enhance the accuracy of recommendation. In the meanwhile, we apply the proposed algorithm into social networks and combine it with trust, thus to address data sparsity and cold start problems to some extent.First, we analyze the advantages and limitations of traditional similarity methods for recommendation algorithm, including Jaccard, Cosine similarity and Pearson Correlation Coefficient. On this basis, in order to further improve the accuracy of recommendation, we exploit different combination methods to integrate Jaccard with Cosine similarity from the perspective of both co-rated items and unpopular items into new algorithms.Second, aiming at the trust relationships among users in social networks, we introduce the trust propagation to extend trust relationships, and refine the original binary users trust relationships to be more accurate. Based on these, we incorporate the refined trust matrix into the global information based recommendation algorithm to resolve data sparsity and cold start problems, thus to further enhance recommendation accuracy.Lastly, to evaluate the practicability and effectiveness of our proposed algorithms,we conduct comprehensive and extensive experiments on two real-world datasets for both the global information based recommendation algorithm and the social network based recommendation algorithm.