基于多社交关系的社团划分概率矩阵推荐算法

Community partition probability matrix recommendation algorithm based on multiple social relations

随着大数据的不断发展,用户的个性化推荐得到普遍应用,现有的推荐算法忽略了用户之间的多种社交关系组成的社团结构,但在现实的网络空间中用户间的多种社交关系可以很好的作用于推荐系统。基于多子网复合复杂网络模型,利用多种社交关系组成的社团结构特性,提出了基于多社交关系的社团划分概率矩阵推荐算法。通过在真实数据集Epinions上与现有推荐算法进行对比,准确率评价指标δMAE、δRMSE分别提高了30%、20%,由此可以证明,基于多社交关系的社团划分概率矩阵推荐算法能有效提高推荐准确率。

With the continuous development of big data,the personalized recommendation of users is widely used.The existing recommendation algorithms ignore the community structure composed of various social relations between users.Relationships can be used well in recommendation systems.Based on the multi-subnet complex network model,this paper proposes a community partitioning probability matrix recommendation algorithm based on multiple social relationships by using the community structure characteristics composed of multiple social relationships.By comparing with the existing recommendation algorithms on the real data set Epinions,the accuracy evaluation indicatorsδMAE andδRMSE have been improved by 30%and 20%respectively.It can be proved that the community division probability matrix recommendation algorithm based on multiple social relations can effectively improve the recommended accuracy.