Matrix Factorization with Explicit Trust and Distrust Relationships

摘要: With the advent of online social networks, recommender systems have became crucial for success many applications/services due to their significance role in tailoring these applications user-specific needs or preferences. Despite increasing popularity, general suffer from data sparsity and cold-start problems. To alleviate issues, recent years there has been an upsurge interest exploiting information such as trust relations among users along with rating improve performance systems. The main motivation recommendation process stems observation that ideas we are exposed choices make significantly influenced by our context. However, large user communities, addition relations, distrust also exist between users. For instance, Epinions concepts personal "web trust" "block list" allow categorize friends based on quality reviews into trusted distrusted friends, respectively. In this paper, propose a matrix factorization model networks properly incorporates both relationships aiming recommendations mitigate issues. Through experiments set, show new algorithm outperforms its standard trust-enhanced distrust-enhanced counterparts respect accuracy, thereby demonstrating positive effect incorporation explicit can