基于高斯分布建模的序列推荐算法  

Algorithm of sequential recommendation based on Gaussian distribution modeling

作　　者：欧道源 梁京章[1] 吴丽娟 Ou Daoyuan;Liang Jingzhang;Wu Lijuan(College of Electrical Engineering,Guangxi University,Nanning 530004,China;Information&Network Center,Guangxi University,Nanning 530004,China)

机构地区：[1]广西大学电气工程学院,南宁530004 [2]广西大学信息网络中心,南宁530004

出　　处：《计算机应用研究》2023年第4期1108-1112,共5页Application Research of Computers

基　　金：广西重点研发计划项目(桂科AB22035033)。

摘　　要：目前大多数序列推荐系统(SRS)都假设需要预测的下一项与用户的上一项输入有关,然而真实场景中,用户可能会在浏览过程中误点击与自身兴趣偏好不一致的项目(不可靠实例)。针对此问题,提出了一种基于高斯分布建模的序列推荐算法。该算法首先通过一个含有多头自注意力的不确定性感知图集合网络(uncertainty-aware graph ensemble network,UAN),通过降低输入项的不确定性来提取输入的序列模式;其次将提取的输入序列模式建模为一个高斯分布,得到序列信息中的动态用户偏好以及偏好的不确定性;再将传统的推荐目标函数拓展为一个采样损失函数和一个不确定性正则化器,赋予每个训练实例适当的不确定性;最后将高损失且低不确定性的不可靠实例去除,增强序列推荐的准确性。该算法在三个公开的数据集Book-Crossing、MovieLens-1M和Steam上进行实验测试,结果表明,该算法相对于效果较好的基线取得了5.3%左右的提升,得到了更优的序列推荐结果,并能通过有效降低输入序列信息的不确定性,从而提升推荐准确率。Most sequential recommendation systems(SRS)assume that the next item to be predicted is related to the user’s previous input.However,in real scenarios,users may click items inconsistent with their own interests and preferences(unreliable instances)by mistake during browsing.This paper proposed a sequential recommendation algorithm based on Gaussian distribution modeling to solve this problem.Firstly,the algorithm extracted the input sequence patterns by reducing the uncertainty of input items through an uncertainty aware graph ensemble network(UAN)with multiple heads of self-attention.Secondly,it modeled the extracted input sequence pattern as a Gaussian distribution,and obtained the dynamic user’s prefe-rences and the uncertainty of preferences in the sequence information.Then,it extended the traditional recommended objective function to a sampling loss function and an uncertainty regularizer,and gave each training instance appropriate uncertainty.Finally,it removed the unreliable examples with high loss and low uncertainty to enhance the accuracy of sequence recommendation.It tested the algorithm on three open datasets,Book-Crossing,MovieLens-1M and Steam.The results show that the algorithm has achieved an improvement of about 5.3%compared with the baseline with good effects,and has obtained better sequential recommendation results.The proposed algorithm can improve the recommendation accuracy by effectively reducing the uncertainty of input sequence information.

关 键 词：推荐系统 序列推荐 高斯分布 不确定性感知 推荐损失 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]