采用信任网络增强的协同过滤算法

摘要

由于数据稀疏性问题的普遍存在,不仅传统的协同过滤系统中使用单一相似度进行的推荐不具备较高的可信度,而且共同评分项过于稀疏也会导致其推荐性能大打折扣.针对以上问题,提出了一种采用信任网络增强的协同过滤算法(记为ECFATN).通过引入社会网络中常用的信任关系,即在原始的用户—项目评分矩阵上,通过信任计算建立用户间的信任关系,并使用传播规则传递信任关系,构建一个用户信任网络;最终使用用户间的信任度与相似度线性加权作为新的权重进行推荐.在真实的数据集上进行测试,实验结果表明,ECFATN算法不仅在一定程度上缓解了数据稀疏性问题并提高了推荐精度,而且由于信任关系的引入,对于用户冷启动问题也有较大的改善.%Due to the widespread data sparsity,traditional collaborative filtering system with single similarity cannot recommend accountably,meanwhile,the sparsity of co-rated items can lead to poor recommendation performance.In view of the above problems,this paper came into consideration an enhanced collaborative filtering algorithm adopting trust network (referred as ECFATN).This algorithm built a users' trust network based on the original user-item rating matrix by introducing the common trust relations of social network,established trust relationships between users with trust computation,and deliveried trust-relations with propagation rules.Then it linearly weighed trust and similarity between users to get a new weight.Finally a new recommendation way came out with this new weight.Experimental results show that this algorithm moderately alleviates data sparsity,and improves recommendation accuracy.Also,with the introduction of trust relations,this algorithm can improve user's cold-start problem.