基于注意力机制的用户动态兴趣推荐算法  

User Dynamic Interest Recommendation Algorithm Based on Attention Mechanism

作　　者：刘星宇 谢颖华[1] 

机构地区：[1]东华大学信息科学与技术学院,上海

出　　处：《计算机科学与应用》2022年第2期269-279,共11页Computer Science and Application

摘　　要：传统推荐算法由于缺乏对用户兴趣的精准捕捉,会导致算法性能不佳等问题。随着深度学习在推荐系统领域应用的不断深入,针对该问题有了许多很好的解决方法。为实现对用户动态兴趣的捕捉与行为的预测,提出一种基于注意力机制的用户动态兴趣推荐算法(Attention Mechanical Recom-mendation Algorithm Based on Long Short-Term, AM_LST)。该算法通过潜在因子模型与门控循环单元,分别实现对用户长短期兴趣的动态捕捉;并使用注意力机制将用户长短期兴趣有机结合,来改善传统推荐算法因用户兴趣的动态变化而导致推荐效果下降的问题。最终通过对比实验,证明了本文提出的模型在性能上的提升。Due to the lack of accurate capture of user interest, the traditional recommendation algorithm will result in poor performance. With the application of deep learning in the field of recommendation system, there are many good solutions to this problem. In order to realize the capture of user’s dynamic interest and behavior prediction, a users’ dynamic interest recommendation algorithm based on attention mechanism (Attention Mechanical Recommendation Algorithm Based on Long Short-Term, AM_LST) is proposed. In this algorithm, latent factor model and gated recurrent unit are used to capture the user’s long short-term interests. In addition, attention mechanism is used to organically combine users’ long and short interests to improve the problem that the traditional recommendation algorithm’s recommendation effect decreases due to the dynamic change of users’ interests. Finally, the performance improvement of the proposed model is proved through comparative experiments.

关 键 词：推荐算法 注意力机制 潜在因子模型 循环神经网络 门控循环单元 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]