提示
用户认可度增强的序列推荐算法
首发时间：2025-03-14
摘要：针对现有序列推荐方法忽视用户认可度信息的问题,提出一种用户认可度增强的序列推荐算法模型uaesrec(user approval-enhanced sequential recommendation),旨在通过挖掘用户对物品的积极/消极评价信息,提升推荐系统的准确性与可解释性。该模型首先构建共认可知识图谱,利用图嵌入技术增强物品语义表示；其次设计双分支注意力网络分别建模群体偏好共识与潜在反感表示；最后引入对比学习机制约束用户表征空间的区分性。在三个公开数据集上的实验结果表明,uaesrec显著优于现有的基线模型,实验证明用户认可度信号能有效揭示群体偏好特征,双分支对比学习机制可显著增强动态偏好建模能力,为解决数据稀疏性和冷启动问题提供了新思路。
关键词：
for information in english, please click here
user approval-enhanced sequential recommendation
abstract：to address the issue that existing sequential recommendation methods neglect user approval information, this study proposes a user approval-enhanced sequential recommendation (uaesrec) model, aiming to improve the accuracy and interpretability of recommendation systems by mining users' positive/negative feedback on items. the model first constructs a co-approval knowledge graph to enhance item semantic representations through graph embedding. then, a dual-branch attention network is designed to model group preference consensus and potential aversion patterns separately. finally, a contrastive learning mechanism is introduced to regularize the representation space. experimental results on three public datasets demonstrate that uaesrec significantly outperforms existing baseline models. the study proves that user approval signals can effectively reveal group preference characteristics, and the dual-branch contrastive learning mechanism significantly enhances dynamic preference modeling capabilities, providing new insights for addressing data sparsity and cold-start problems.
keywords：
基金：
论文图表：
引用
导出参考文献
no.****
同行评议
勘误表
用户认可度增强的序列推荐算法
评论
全部评论