SBHCF:基于奇异值分解的混合协同过滤推荐算法

SBHCF: An SVD-based Hybrid Collaborative Filtering Recommendation Algorithm

针对传统协同过滤中的最近邻查找不够合理导致推荐的准确率较低的困境。提出一个基于矩阵分解的混合相似度算法。该方法融合了基于模型的奇异值矩阵分解算法和基于近邻的协同过滤算法皮尔逊相关系数,并引入阈值和杰卡德系数对相似度进行修正。在公共有效数据集上的实验表明,所提出算法的平均绝对误差比传统的推荐算法至少降低了7.7%,有效提高了推荐准确率。

The traditional CF recommendation algorithms are poor in accuracy because of its irrational neighborretrieve. In this paper,an SVD-based hybrid collaborative filtering algorithm is proposed to solve the challenge.The method combines SVD model-based CF algorithm and PCC memory-based CF algorithm. Several parameters and JACCARD are introduced to revise the similarity. The experiment in the public data set proves that the SBHCF algorithm effectively improves the recommendation accuracy with a reduced MAE by at least 7. 7% than the traditional CF algorithm.