基于神经网络和注意力机制的协同过滤推荐算法的研究

Research on collaborative filtering recommendation algorithm based on neural network and attention mechanism

针对协同过滤推荐算法中用户-物品矩阵的稀疏性,使得传统协同过滤算法推荐度较差的问题,提出一种改进的基于神经网络和注意力机制的协同过滤推荐算法B-SDAECF,旨在解决传统推荐系统中数据稀疏的问题。结合Transformer模型的变式Bert模型和堆叠式降噪自动编码器(SDAE),利用Bert模型从用户评论中提取高质量的特征表示,以获得向量矩阵;并将向量矩阵作为SDAE的初始权重,从而使SDAE模型能够更快速地运算,进而填充原有的用户-项目评分矩阵。实验结果显示,相比传统方法,所提方法在推荐系统的准确性和鲁棒性上有显著提升,推荐效果更优秀。

In order to solve the problem that the sparseness of the user-item matrix in the collaborative filtering recommendation algorithm can cause poor recommendation degree of the traditional collaborative filtering algorithm,an improved collaborative filtering recommendation algorithm B-SDAECF based on neural network and attention is proposed to solve the problem of data sparsity in the traditional recommendation system.By combining the transformer Bert model of the Transformer model and the stacked denoise auto-encoder(SDAE),the Bert model is used to extract high-quality feature representations from user reviews to obtain the vector matrix.The vector matrix is used as the initial weight of the SDAE,so that the SDAE model can be operated more quickly,and then the original user-item scoring matrix is filled.The experimental results show that,in comparison with the traditional method,the proposed method can significantly improve the accuracy and robustness of the recommendation system,and has better recommendation performance.