优化的协同过滤推荐算法

Optimized collaborative filtering recommendation algorithm

针对传统协同过滤中存在的数据稀疏性和用户偏见性问题进行研究,提出一种基于用户模型的相似度计算方法。通过建立用户偏好主题向量,间接获取用户之间的相似性,克服数据稀疏性对用户相似度计算准确性的影响;基于对用户偏见性的考虑,引入贝叶斯重排序算法,建立项目的信任子群,获取用户对信任子群的局部偏见,通过加权相似用户对目标项目的评分产生推荐。相关实验验证了该方法的可行性,能在有效克服传统协同过滤方法弊端的基础上,提升系统的推荐效果。

Aiming at the problems of data sparsity and users＇ prejudice of traditional collaborative filtering,a method for similarity calculation based on the user model was proposed.User preferences topic vector was established and the similarity between the users was gained indirectly,the negative effects of data sparsity on accuracy in users＇ similarity measure were overcome.Based on the consideration of users＇ prejudice,trusted subgroup of each item was constructed using Bayesian reranking algorithm,the users＇ local biases about the neighbours in subgroup were got.Recommendation was generated by weighting scores of the target coming from similar users.Experimental results indicate that the proposed algorithm can effectively avoid the disadvantages of traditional methods and improve the system＇s recommendation quality.