一种非对称相似度矩阵约束的群组协同过滤算法

Asymmetric Similarity Matrix for High Scalability in Group-based Collaborative Filtering

针对传统协同过滤算法无法处理社交网络中逐渐形成的以群组为中心的群体行为且很少考虑用户间普遍存在的不对称关系,提出一种非对称因子约束相似度矩阵的群组协同过滤算法.首先通过构建虚拟用户作为群组特征,把多维数据降维成与用户特征等价的虚拟用户,以替代群组进行相似度计算.其次引入影响因子和偏移因子两种相似度限制因子来构建非对称相似矩阵,影响因子表示用户相互影响力的大小关系,偏移因子则考虑用户评分习惯间的差异.实验结果表明该方法与传统推荐算法相比具有收敛速度快、预测精度高的优势,提高了推荐质量并解决了群组特征处理问题.

The traditional collaborative filtering algorithm can＇t deal with the group behavior in social networks and rarely consider the asymmetric relationship of users. The paper proposes a group collaborative filtering algorithm with asymmetric factor constraints based on virtual user. Firstly, constructing the virtual user, same as user feature, instead of the group feature to transform user-item score ma- trix to virtual user-item score matrix and compute similarity matrix. Then, the interaction of influence factor and tendency factor adjust adaptively item to acquire asymmetric similarity matrix. The influence factor indicates the size of user＇s influence, and tendency factor is the difference between the user habits. Experimental results show that the proposed algorithm achieves better result on the accuracy of recommendation and solves the problem of how to process group feature.