Abstract:
Aiming at the problem of sparse user ratings facing tradition recommendation system, this paper proposes a hybrid recommendation method based on Tag and collaborative filtering(TAG-CF) to provide a solution to this problem. The neighbors set for the target item can be gained based on tagging information. It uses item-based collaborative filtering to generate the predictive ratings. By filing these predictive ratings into the sparse user-item rating matrix, it constructs a full pseudo ratings matrix. It computes the predictions based on the pseudo ratings matrix by using user-based collaborative filtering. Experimental results show that the proposed method performs significantly better than the traditional CF method.
Key words:
recommendation system,
collaborative filtering,
sparsity,
mass taxonomy,
Tag
摘要:
针对传统协同过滤方法的稀疏性问题,提出基于标签(Tag)和协同过滤的混合推荐方法TAG-CF。通过Tag分类信息获取项目的最近邻居,采用基于项目的最近邻方法预测用户评分值,并利用该预测值填充用户评分矩阵,构造密集的伪矩阵,运用基于用户的的协同过滤方法在伪矩阵上计算用户对项目的预测评分值。实验结果表明,TAG-CF能有效降低推荐系统的平均绝对误差,提高推荐质量。
关键词:
推荐系统,
协同过滤,
稀疏性,
大众分类法,
标签
CLC Number:
WANG Wei-Beng, WANG Jin-Hui. Hybrid Recommendation Method Based on Tag and Collaborative Filtering[J]. Computer Engineering, 2011, 37(14): 34-35.
王卫平, 王金辉. 基于Tag和协同过滤的混合推荐方法[J]. 计算机工程, 2011, 37(14): 34-35.