摘要： 传统的协同过滤推荐算法存在数据稀疏和可用用户偏好信息有限的问题。针对数据稀疏问题和联系不可靠现象对传统社交网络推荐带来的影响,提出一种在信任关系重建和社交网络传递基础上的推荐算法。引入去伪存真方法避免联系不可靠现象,根据用户所建立的联系规 模改进用户相似度计算公式提高去伪存真准确度。定义预备朋友的概念,为用户推荐预备朋友解决数据稀疏问题。在重建的信任关系上利用社交网络的传递性进行预测评分,并定义二端分布函数调节预测评分提高推荐准确度。在数据集Epinion上进行实验,结果表明,该算法 能减小数据稀疏和联系不可靠现象对推荐结果的负面影响,有效降低预测结果的平均绝对误差,提高推荐的准确度。

关键词: 信任关系, 社交网络, 预备朋友, 二端分布函数, 推荐算法

Abstract: The challenge of traditional collaborative filtering recommendation algorithms is data sparsity and limited user-preference information that is available.With the rapid development of the Internet,users can take the initiative to establish contact with other users who are trustworthy and have similar ideas.Some users may establish links which are unreliable.To overcome the impact of data sparsity and unreliable links on traditional social networking recommen-dations,this paper presents a recommendation algorithm based on trust relationships rebuilding and social networks transferring.It uses the method,called discarding the false and retaining the true,to solve the phenomenon of unreliable links.Meanwhile,it uses the improved jaccadr coefficient based on the scale of link to improve the accuracy of discarding the false and retaining the true.It defines the concept of probationary friends to reduce problem of data sparsity.It uses the transitivity of social network based on the new rebuilded social network to predict and define the concept of two-terminal distribution function and to adjust the predict scoring and improve the predict accuracy.The proposed algorithm is experimented in real Epinions datasets.Experimental results show that the algorithm can reduce the negative impact of the phenomenon of unreliable links and data sparsity,effectively lower the mean absolute error of prediction consequences,and enhance recommendation precision.

Key words: trust relationship, social network, probationary friend, two-terminal distribution function, recommen-dation algorithm

中图分类号: 

• TP18