基于网络表示学习与深度学习的推荐算法研究

Research on Recommendation Algorithm Based on Network Representation Learning and Deep Learning

针对传统基于协同过滤的推荐算法信息提取能力有限的问题,提出基于网络表示学习的卷积协同过滤推荐算法。将二分网络分成物品与用户同质网络,在各自的同质网络上使用GraphSAGE模型得到融合网络空间信息和用户与物品属性信息的矩阵。在此基础上,利用外积运算丰富用户和物品特征向量各维度的相关表示,通过卷积神经网络训练物品和用户的交互信息得到算法模型。实验结果验证了该算法的有效性,且相比ConvNCF算法,其在Movielens数据集上HR@5和NDCG@5分别提升了1.89和2.19个百分点,在Last.fm数据集上HR@5和NDCG@5分别提升了1.09和2.32个百分点。

The traditional recommendation algorithms based on collaborative filtering information are faced with the bottleneck of the information extraction capability.To address the problem,a recommendation algorithm based on convolution collaborative filtering for network representation learning is proposed.The algorithm decomposes the bipartite network into a homogenous network of users and items,and then uses the GraphSAGE algorithm for the two homogenous networks to obtain feature vectors.On this basis,the outer product operation is performed to enrich the relevant representation of each dimension of the user and item feature vectors.Finally,the convolutional neural network is used to train the interaction information between the item and the user to obtain the algorithm model.The experimental results show that the algorithm is effective,compared with the ConvNCF algorithm,the proposed algorithm improves HR@5 by 1.89 percentage point and NDCG@5 by 2.19 percentage point on the Movielens data set.The improvement of HR@5 is 1.09 percentage point and that of NDCG@5 is 2.32 percentage point on the Last.fm data set.