融合评分矩阵和评论文本的Deep-FRR评分预测模型  

Deep-FRR Rating Prediction Model Combined Rating Matrix and Review Text

作　　者：侯越[1] 谢斌 陈佳兴 HOU Yue;XIE Bin;CHEN Jia-xing(School of Electronic and Information Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]兰州交通大学电子与信息工程学院,兰州730070

出　　处：《兰州交通大学学报》2020年第5期48-54,共7页Journal of Lanzhou Jiaotong University

摘　　要：基于矩阵分解的评分预测模型可有效解决传统推荐算法中评分矩阵的稀疏性问题,但此预测模型未考虑评论文本的用户喜好和项目特征对评分结果的影响.为提高评分预测精度,深入挖掘评论文本潜在的语义特征,提出一种由矩阵分解模块、深度矩阵分解模块以及卷积神经网络模块构建的Deep-FRR评分预测模型,矩阵分解模块用于学习用户和项目间的线性交互关系,深度矩阵分解模块用于学习用户和项目间的非线性交互关系,卷积神经网络模块用于提取用户、项目评论文本的深层特征表达,基于全连接层实现三模块的融合,然后进行评分预测.实验结果表明,Deep-FRR评分预测模型在不同测试集上的均方误差均低于其它对比模型.Rating prediction model based on matrix factorization effectively solves the sparsity problem of traditional collaborative filtering algorithm.However,this prediction model does not take the influence of user preferences and item characteristics of the comment text on the scoring results into consideration.In order to further enhance the precision of rating prediction and explore the potential semantic features of comment text,this paper proposed a Deep-FRR model which consists of three main modules,in which the matrix factorization and deep matrix factorization modules can extract the linear and non-linear interactions of users and items,while the convolutional neural network module can extract the deep semantic expression of users and items from rating texts.Then the three modules were merged in the full connected layer to produce the final rating prediction result.The experimental results show that the mean square error of the proposed model on the test set is prominently lower than that of other models.

关 键 词：推荐系统 矩阵分解 卷积神经网络 评分预测 模型融合 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]