This paper gives a detailed analysis of the existing probabilistic matrix factorization algorithms, and proposes a social recommendation algorithm based on user similarity (SRUS) for solving the problem of inaccurate predictions based only on the rating information. Firstly, on the basis of the probabilistic matrix factorization (PMF) algorithm, its modeling is built on the information of user similarity. Secondly, it uses the potential feature space to integrate the rating matrix and similarity matrix into a unified framework. Finally, the matrix factorization of these two matrices is carried out to optimize the algorithm. The comparison of this algorithm with the PMF algorithm shows that this algorithm is better in data sparsity, cold start and accuracy.%在对现有的概率矩阵分解算法研究的基础上, 针对其中只使用评分信息来做预测存在较大误差的问题, 提出了一种结合用户相似度的社会化推荐算法 (SRUS).首先, 以概率矩阵分解算法 (PMF) 为基础, 结合用户相似度信息进行建模;其次, 使用潜在特征空间将评分矩阵和相似度矩阵关联到一个统一的框架中;最后, 对这2个矩阵进行矩阵分解, 实现算法的优化推荐.将这一算法与PMF算法进行比较, 实验表明, SRUS算法在数据稀疏性、冷启动和精确性方面具有更优的效果.
展开▼
