融合信任关系与评论文本的矩阵分解推荐算法

针对推荐系统中用户评分数据稀疏所导致推荐结果不精确的问题,本文尝试将用户评分、信任关系和项目评论文本信息融合在概率矩阵分解方法中以缓解评分数据稀疏性问题.首先以共同好友数目及项目流行度改进皮尔逊用户偏好相似程度并将其作为用户间的直接信任值,然后考虑用户间信任传播过程中所有路径的影响构建新的信任网络;其次通过BERT预训练(Pre-training of Deep Bidirectional Transformers for Language Understanding)模型提取项目的评论文本向量,构造项目的评论文本特征矩阵;最后基于概率矩阵分解(Probabilistic Matrix Factorization,PMF)模型融合用户的评分数据、用户的信任关系以及项目的评论文本信息进行推荐.通过不断的理论分析并在真实的Yelp数据集上进行实验验证,均表明本文算法的有效性.

Matrix Factorization Recommendation Algorithms by Exploiting Trust Relationship and Review Text

Aiming at the problem of inaccurate recommendation results caused by sparse user rating data in the recommendation system,this article attempts to fuse user rating,trust relationships,and item review text information into a probability matrix decomposition algorithm to alleviate the problem of sparse rating data.First,improve the similarity degree of Pearson user preferences with the number of common friends and project popularity as the direct trust value between users,and then consider the impact of all paths in the process of trust propagation between users to build a new trust network;secondly,through BERT pre-trained(Pre-training of Deep Bidirectional Transformers for Language Understanding)model extracts the content of the review text of the project and constructs the potential features of the review text of the project;finally,the user's rating data is integrated based on the Probabilistic Matrix Factorization(PMF)model,The user's trust relationship,and the item's comment text information.Through continuous theoretical analysis and experimental verification,the effectiveness of the algorithm in this paper is demonstrated.