欢迎访问《图学学报》 分享到：
图学学报

图学学报

• 专论：第16届媒体智能与大数据计算会议（CIDE & DEA 2019 大连） • 上一篇    下一篇

基于协同回归模型的矩阵分解推荐

  

  1. (1. 中国农业大学信息与电气工程学院，北京 100083；
     2. 北京市农业物联网工程技术研究中心，北京 100083； 
    3. 农业农村部农业信息获取技术重点实验室，北京 100083； 
    4. 鲁东大学信息与电气工程学院，山东 烟台 264025)
  • 出版日期:2019-12-31 发布日期:2020-01-20
  • 基金资助:
    国家自然科学基金项目(61472172)

Matrix Factorization Recommendation Based on  Collaborative Regression Model

  1. (1. College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, China;  
    2. Beijing Agricultural Internet of Things Engineering Technology Research Center, Beijing 100083, China; 
    3. Key Laboratory of Agricultural Information Acquisition Technology of the Ministry of Agriculture and Rural Affairs, Beijing 100083, China; 
    4. College of Information and Electrical Engineering, Ludong University, Yantai Shangdong 264025, China)
  • Online:2019-12-31 Published:2020-01-20

PDF (PC)

128

摘要： 推荐系统是解决信息过载的有效途径。传统的推荐系统难以从海量数据中推选出 符合用户个性化偏好的项目，推荐质量不高。为此，通过优化传统的协同过滤推荐算法，针对 数据稀疏性等问题，提出协同回归模型的矩阵分解算法(CLMF)。通过机器学习算法发掘内容信 息的深层次特征，提升了原始数据的信息量；并构建辅助特征矩阵，通过融合特征矩阵，CLMF 最大化了特征标签的作用，并结合数据标签，语义信息和评分矩阵得到推荐算法框架。在真实 数据集上实验结果显示，新型推荐算法可有效解决特征值缺失问题，改善了数据稀疏性，提升 了算法扩展性，并显著增强覆盖性。

关键词: 推荐系统, 协同过滤, 机器学习, 矩阵分解, 稀疏性, 覆盖性

Abstract: Recommendation system is an effective way to solve the problem of information overload. It is difficult for the traditional recommendation system to select items that meet the user’s personalized preferences from mass data, and the recommendation quality is disappointing. By optimizing the traditional collaborative filtering recommendation algorithm, this paper proposes a matrix factorization algorithm (CLMF) for collaborative regression model in view of data sparsity and other problems. The method uses machine learning algorithm to identify the in-depth characteristics of content information, which increases the information capacity of original data. Constructs auxiliary feature matrix, maximizes the role of feature labels through feature matrix fusion, and combines data labels, semantic information and scoring matrix to get the algorithm framework. The experimental results on the real dataset show that the new recommendation algorithm can effectively solve the problem of missing eigenvalues, improve the data sparsity, significantly enhance the scalability of the algorithm and the coverage.

Key words: recommendation system, collaborative filtering, machine learning, matrix factorization, sparsity, coverage