Abstract: Traditional friends recommendation algorithms only consider the topological similarity in calculating the similarity of friends.The similarity of users’ interests is seldomly taken into account,and the recommendation results are often not precise enough.Many existing social networking sites (such as Douban.com) provide the functions of user ratings,i.e. users can give ratings for certain types of items (such as movies).In order to calculate user’s interests similarity,we proposed a method which computes the interests similarity based on the ratings given by users and got more accurate result of recommendation by incorporating the interests similarity into the topological similarity.Firstly,we used cosine similarity to calculate the topological similarity between users.In calculating the interests similarity based on user ratings,we got the users cluster rating similarity matrix through the establishment of a probabilistic model,and derived users’ interests similarity from the rating similarity matrix.Finally,users’ interests similarity and topological similarity were combined to get the final improved friends recommendation algorithm.In order to verify the effectiveness of our method,we applied our method to the crawled user data from Douban website.The experimental results show that our method can effectively improve the accuracy of recommendation results compared with the traditional recom-mendation algorithm based on topology similarity.

Key words: Social network,Recommendation,Topology,Rating,Cluster,Similarity