Research On The Sparsity Problem Of Recommender Systems Based On Social Networks

With the development of information technology and information resources expanding rapidly, it is more difficult for people to get the information that they are really interested in, which is well-known information overload. Recommender systems are regarded as an important approach to dealing with information overload, and it can predict users’ preferences according to the known users’records and then help users gain the information which they are interested in. Recommender systems have been used in many fields, such as electronic commerce, personalized advertisement recommendation, personalized online services, etc.At present, collaborative filtering is the most successful approach, but it faces the terrible sparsity problem. The two main approaches to handling the sparsity problem are improving the density of rating matrix and amending the key recommending processes. This paper deals with the sparsity problem from the perspective of adding kinds of information, and then adds friendships,tags, and neighbors of item into recommending models in order to reduce the impacts of the sparsity problem. In research process, this paper firstly presents the sparsity problem of the ratings, and then analyzes the impacts of users’ friends and tags on reducing the sparsity problem, and then analyzes the impact of neighbors of item on reducing the sparsity problem. Finally, this paper validates the impacts of users’friends, tags and neighbors of item on reducing the sparsity problem according to the experiments.