【摘要】 提出一种嵌入用户评分偏好置信度的社会化推荐算法Conf-SMF,将用户评分偏好与拓展后的社会信任关系结合起来,有效提高了推荐质量。在Film Trust、CiaoDVD和Epinions3个公开数据集上进行实验,实验结果表明:提出的算法相比Trust MF、CUNE-MF推荐算法,在3个数据集上预测误差最大分别降低5.79%、4.58%; 14.13%、12.84%; 10%、8.77%。另外,所提出的算法对"冷启动"用户与"活跃"用户的预测评分性能也有所提高。更多还原

【Abstract】 A social matrix factorization recommendation algorithm with user rating preference confidence conv SMF is proposed,which combines user rating preference with user rating preference confidence,and effectively improves the quality of recommendation. Experiments are carried out on film trust,ciaodvd and epinions. The experimental results show that,compared with trustmf and cune-mf recommendation algorithms,the prediction errors of the proposed algorithm are reduced by 5.79%,4. 58%,14. 13%,12. 84%,10% and 8. 77%,respectively. In addition,the performance of the proposed algorithm for "cold start"users and "active"users is also improved.更多还原