基于最大团的社交网络个性化推荐

Personalized Recommendation of Social Network Based on Maximum Clique

传统的协同过滤算法理论上是基于社会理论的,因为它们与我们参考其他人的选择的过程类似。但比较严重的冷启动问题一直影响着协同过滤算法的实际效率。论文基于传统的协同过滤推荐算法,以用户社交网络中的最大团为单位进行团推荐,利用社会化网络中好友圈的兴趣汇聚及相互影响,通过团好友对物品的评分的平均数来作为作为团成员的推荐分。实验证明了协同过滤中使用最大团与相似的人混合推荐而不只是使用相似的人进行推荐,不仅可以提升推荐的准确率,而且对于传统推荐系统中比较严重的冷启动问题也得到了一定程度的缓解。

Traditional collaborative filtering algorithms are theoretically based on social theory because they are similar to the way we refer to other people's choices. However,the serious cold boot problem has always affected the practical efficiency of collab. orative filtering algorithm. The traditional collaborative filtering recommendation algorithm is based on the group recommended to the largest group of users in a social network as a unit,the interaction and convergence of interest friends in social network,the aver. age number of items in the group of friends as members of the score points recommended. The experiment proves that collaborative filtering using the maximum clique and people with similar mixed recommendation and not just the use of similar people recommend. ed not only can improve the accuracy of recommendation,but also to some extent alleviate the cold start problem is more serious in traditional recommendation system.