针对传统协同过滤算法中用户交叉评分项较少的情况,提出一种相似度与社交网络中信任因子结合的新方法.首先利用评分矩阵获取用户间相对缺失的评分项.其次通过概率矩阵分解技术进行降维获得近似评分矩阵,用以动态填充上述缺失项目评分,以填充后的用户评分计算用户间相似度.对于部分填充中存在误差的项目通过信任因子动态调整,获得更符合实际的相似度.在公开数据集MovieLens的实验结果显示,该方法较传统方法在推荐精度方面提升2.1%以上.%Since there are few users cross rating items in the traditional collaborative filtering algorithm,a new method is proposed to combine the similarity with the trust factor in the social network.First,the rating matrix is adopted to obtain the relative missing rating items between the users.Then,an approximate rating matrix is obtained by probabilistic matrix factorization method to selectively fill the missing rating matrix.Meanwhile,the trust factor is presented to adjust the error in process of calculate similarity.The results on the publicly available MovieLens datasets show that the proposed algorithm can improve the recommendation accuracy by above 2.1% on the classic algorithm.
展开▼