基于用户综合信任度与社区信任传播的推荐算法

Recommendation Algorithm Based on User Comprehensive Trust Degree and Community Trust Propagation

传统的协同过滤推荐算法存在数据稀疏性、用户冷启动等问题,基于信任机制的推荐算法虽然能够缓解数据稀疏性问题,但是在信任传播过程中时间成本过高。为此,提出基于用户综合信任度与社区信任传播的推荐算法,通过算法中的虚拟社区信任模型获取用户综合信任度,将其带入协同过滤算法得到推荐结果。该算法综合考虑显性和隐性2种直接信任度,得到直接综合信任度构建用户信任网络,并利用重叠社区发现算法为用户划分专属虚拟社区进行信任传播,弥补直接综合信任度数量的不足。在Epinions数据集上的实验结果表明,该算法能够在缓解数据稀疏性和用户冷启动问题的同时,降低信任传播的时间成本,提高推荐质量。

The traditional collaborative filtering recommendation algorithms have the problems of data sparsity and cold start of users.Although the recommendation algorithms based on trust mechanism can alleviate the problem of data sparsity,but the time cost is large in the process of trust propagation.Aiming at this problem,a recommendation algorithm based on user comprehensive trust and community trust propagation is proposed in this paper.The trust model of this algorithm is used to obtain comprehensive trust which is brought into collaborative filtering algorithm to get the recommendation results.The proposed algorithm considers two direct trust of explicit and implicit,and gets direct comprehensive trust which is used to build trust network of users.Private virtual community for users is divided by using the overlapping community discovery algorithm,and trust propagation is carried out in the user private virtual community to compensate for the lack of direct comprehensive trust.The experimental result on Epinions data set shows that this algorithm can effectively alleviate the data sparsity and cold start of users,meanwhile reducing the time cost of trust propagation and improving the quality of recommendation.