许多传统的推荐方法如协同过滤和低秩矩阵分解都存在物品或用户方面的稀疏性和冷启动问题.为了克服这两方面的问题,提出一种基于隐语义模型的个性化推荐方法.通过对用户行为进行分析,利用隐语义模型推断出用户潜在的兴趣因子,从而构建用户兴趣特征矩阵来进行个性化推荐.对现实的电影数据的实验证明了所提方法的有效性,并在准确率、召回率和覆盖率方面均优于传统的协同过滤方法和基于内容的方法.%Many traditional recommendation methods,such as collaborative filtering or low rank matrix factorization,have data sparsity and cold-start problems on items or users.In order to overcome these two problems,this paper proposes a personalized recommendation method based on the latent factor model.By analyzing behaviors of users,we utilize latent factor model to infer user interest feature vector so as to make a personalized recommendation.Our two kinds of experiments on realistic movie data demonstrate the efficacy of the proposed method,as well as the superiority compared to traditional collaborative filtering methods and content-based methods.
展开▼
