融合元学习和注意力机制的跨域推荐算法研究  

Research on Cross-Domain Recommendation Algorithm Integrating Meta-Learning and Attention Mechanism

作　　者：郭佳 郑山红[1] 陈闯 王国春[1] GUO Jia;ZHENG Shan-hong;CHEN Chuang;WANG Guo-chun(School of Computer Science and Engineering,Changchun University of Technology,Changchun Jilin 130012,China)

机构地区：[1]长春工业大学计算机科学与工程学院,吉林长春130012

出　　处：《计算机仿真》2024年第12期344-348,共5页Computer Simulation

基　　金：吉林省科学技术厅重点研发项目(20220201159GX)。

摘　　要：冷启动问题是推荐系统面临的一个巨大挑战,跨域推荐是解决冷启动问题的主要方法之一。现有的大多数方法都是在源域和目标域之间建立一个所有用户公用的连接桥,但是用户的需求各不相同,关注的方面也各不相同,一个公用的连接桥不能很好的连接源域和目标域,所以提出了一个新的跨域推荐框架。首先将卷积神经网络和前馈神经网络结合进行特征提取,提高模型的鲁棒性。其次通过元学习来学习源域嵌入,为每个用户生成带有用户特征的个性化连接桥。最后,将注意力机制融入到源域的用户嵌入,提高推荐的准确率。在跨域推荐场景上的实验结果表明,上述模型的性能优于其它跨域推荐模型。The cold start problem is a huge challenge for recommender systems.Cross-domain recommendation is one of the main methods to solve the cold start problem.Most of the existing methods are to establish a connection bridge between the source domain and the target domain,but the requirements of users are different,and the concerns of users are different,so a common connection bridge can not connect the source domain and the target domain well.Therefore,this paper proposed a new cross-domain recommendation framework.Firstly,the convolutional neural network and feedforward neural network were combined for feature extraction to improve the robustness of the model.Secondly,meta-learning was used to learn the source domain embedding,and a personalized connection bridge with user characteristics was generated for each user.Finally,the attention mechanism was integrated into the user embedding of the source domain to improve the accuracy of the recommendation.Experimental results on cross-domain recommendation scenarios show that the performance of the proposed model is better than other cross-domain recommendation models.

关 键 词：跨域推荐 元学习 冷启动问题 注意力机制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]