Computer Engineering and Applications ›› 2021, Vol. 57 ›› Issue (19): 198-204.DOI: 10.3778/j.issn.1002-8331.2006-0012

Previous Articles     Next Articles

Social Recommendation Algorithm by Fusing Item Information

WEI Dingfeng, LI Liang, CHAI Jing   

  1. 1.School of Information and Computer, Taiyuan University of Technology, Taiyuan 030600, China
    2.School of City and Tourism, Taiyuan Normal University, Taiyuan 030619, China
  • Online:2021-10-01 Published:2021-09-29

融合物品信息的社会化推荐算法

卫鼎峰,李梁,柴晶   

  1. 1.太原理工大学 信息与计算机学院,太原 030600
    2.太原师范学院 城市与旅游学院,太原 030619

Abstract:

Most social recommendation algorithms only constrain the user’s feature vector but not the item’s feature vector. Aiming at this problem, a social recommendation algorithm by fusing item information is proposed. The model firstly constructs an item similarity network based on the user-item interaction diagram. Based on this, random walks and SkipGram’s method are used to construct an item similarity network, and learns the item similarity network, social network and the user-item interaction diagram through the graph neural network to obtain the feature vectors of the users and items coding, and finally on the basis of matrix decomposition, further constraints on the feature vectors of the user and items at the same time, and the iterative update method is used to obtain the final feature vectors of the users and items. In order to verify the performance of the recommendation algorithm, experiments are performed on the FilmTrust, Ciao and Douban datasets. The experimental results show that the proposed ISGCF algorithm has a better recommendation effect than the classic recommendation algorithm, and effectively alleviates the cold start problem.

Key words: social network, object similarity propagation, random walk, cold start, recommendation algorithm, graph neural network

摘要:

大多数社会化推荐算法仅考虑约束用户的特征向量并未限制物品的特征向量对推荐系统性能的影响,针对这一问题,提出了一种融合物品信息的社会化推荐算法。该算法先通过用户与物品的交互图构建物品相似性网络,在此基础上采用随机游走和SkipGram的方法构造出隐性物品相似性网络,再通过图神经网络的方法学习物品隐性相似性网络、社交网络和用户物品交互图,得到用户和物品编码的特征向量,最后在矩阵分解的基础上同时对用户和物品的特征向量做进一步约束,采用迭代更新的方式获取用户和物体最终的特征向量。为验证推荐算法的性能,在FilmTrust、Ciao和Douban数据集上进行实验验证。实验结果表明,所提出的ISGCF算法与经典的推荐算法相比,推荐效果更好,有效地缓解了冷启动问题。

关键词: 社交网络, 物体相似性传播, 随机游走, 冷启动, 推荐算法, 图神经网络