基于时频变换与动态图注意力的多模态序列推荐方法

隋欣怡; 王瑞琴; 任宇彬; 方驰

doi:10.11959/j.issn.1000-0801.DXKX250483

您当前的位置：

首页 >

文章列表页 >

基于时频变换与动态图注意力的多模态序列推荐方法

研究与开发 | 更新时间：2026-06-17

- 基于时频变换与动态图注意力的多模态序列推荐方法
- Multimodal sequential recommendation method based on time-frequency transformation and dynamic graph attention
- 电信科学 2026年42卷第5期页码：169-182
- 作者机构：
  
  湖州师范学院信息工程学院，浙江湖州 313000
- 作者简介：
  
  隋欣怡（2000− ），女，湖州师范学院信息工程学院硕士生，主要研究方向为深度学习、自然语言处理、个性化推荐。
  王瑞琴（1979− ），女，博士，湖州师范学院信息工程学院教授，主要研究方向为自然语言处理、社交网络分析、个性化推荐。
  任宇彬（2000− ），男，湖州师范学院信息工程学院硕士生，主要研究方向为个性化推荐、数据挖掘。
  方驰（2000− ），男，湖州师范学院信息工程学院硕士生，主要研究方向为序列推荐、多模态推荐。
- 基金信息：
  
  国家自然科学基金资助项目;The National Natural Science Foundation of China(62277016)
- DOI：10.11959/j.issn.1000-0801.DXKX250483
  中图分类号： TP391.3;TN919.3
- 收稿：2025-07-30，
  
  修回：2025-11-19，
  
  录用：2025-11-19，
  
  纸质出版：2026-05-20
- 稿件说明：
移动端阅览
隋欣怡,王瑞琴,任宇彬等.基于时频变换与动态图注意力的多模态序列推荐方法[J].电信科学,2026,42(05):169-182.

Sui Xinyi,Wang Ruiqin,Ren Yubin,et al.Multimodal sequential recommendation method based on time-frequency transformation and dynamic graph attention[J].Telecommunications Science,2026,42(05):169-182.
隋欣怡,王瑞琴,任宇彬等.基于时频变换与动态图注意力的多模态序列推荐方法[J].电信科学,2026,42(05):169-182. DOI： 10.11959/j.issn.1000-0801.DXKX250483.

Sui Xinyi,Wang Ruiqin,Ren Yubin,et al.Multimodal sequential recommendation method based on time-frequency transformation and dynamic graph attention[J].Telecommunications Science,2026,42(05):169-182. DOI： 10.11959/j.issn.1000-0801.DXKX250483.

摘要

针对多模态序列推荐中用户兴趣演化建模、跨模态语义对齐及时频特征提取的不足，提出小波增强的动态图注意力推荐（wavelet-enhanced dynamic graph attention recommendation，Wave-DGARec）模型。该模型从时频变换、图建模与对比学习这3个维度进行创新设计：引入多尺度小波变换模块，对行为序列进行时频重构，捕捉非平稳偏好波动；构建用户-商品-图像三元动态图，利用图注意力机制实现结构化语义在不同模态间的高效传播；设计跨模态对比学习机制，引入可学习温度参数，提升语义对齐质量与样本判别能力。在Amazon 4个领域数据集的实验验证了Wave-DGARec的优越性。消融实验证实了小波模块与动态图建模的有效性，为多模态推荐系统提供了一种融合时频分析与结构建模的新范式。

Abstract

To address limitations in modeling user interest dynamics

cross-modal semantic alignment

and time-frequency feature extraction in multimodal sequential recommendation

wavelet-enhanced dynamic graph attention recommendation (Wave-DGARec) was proposed. This framework introduced innovations in three dimensions. A multi-scale wavelet transformation module was introduced to reconstruct behavioral sequences in the time-frequency domain

enabling the capture of non-stationary preference fluctuations. A user-item-image tripartite dynamic graph was constructed

wherein a graph attention mechanism is leveraged to enable efficient propagation of structured semantics across different modalities. A cross-modal contrastive learning strategy with a learnable temperature parameter was designed

enhancing both semantic alignment and sample discrimination. Experimental results on four Amazon domain datasets demonstrated the superiority of Wave-DGARec. Ablation studies further validated the effectiveness of both the wavelet module and the dynamic graph modeling. This work introduced a novel paradigm for multimodal recommendation systems by seamlessly integrating time-frequency analysis with structured representation learning.

关键词

Keywords

references

Kumar C , Chowdary C R , Meena A K . Recent trends in recommender systems: a survey [J ] . International Journal of Multimedia Information Retrieval , 2024 , 13 ( 4 ): 41 .

Chen X , Xu H T , Zhang Y F , et al . Sequential recommendation with user memory networks [C ] // Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining . New York : ACM Press , 2018 : 108 - 116 .

Tao Z L , Liu X H , Xia Y W , et al . Self-supervised learning for multimedia recommendation [J ] . IEEE Transactions on Multimedia , 2023 , 25 : 5107 - 5116 .

Zheng Z , Chao W S , Qiu Z P , et al . Harnessing large language models for text-rich sequential recommendation [C ] // Proceedings of the ACM Web Conference 2024 . New York : ACM Press , 2024 : 3207 - 3216 .

Liu Z H , Mei S , Xiong C Y , et al . Text matching improves sequential recommendation by reducing popularity biases [C ] // Proceedings of the 32nd ACM International Conference on Information and Knowledge Management . New York : ACM Press , 2023 : 1534 - 1544 .

Ong R K , Khong A W H . Spectrum-based modality representation fusion graph convolutional network for multimodal recommendation [C ] // Proceedings of the Eighteenth ACM International Conference on Web Search and Data Mining . New York : ACM Press , 2025 : 773 - 781 .

Deng H Y , Hu J L . Improving sequential recommendation with global item transitions and local subsequences [J ] . IEEJ Transactions on Electrical and Electronic Engineering , 2024 , 19 ( 1 ): 100 - 108 .

李家乐 , 王瑞琴 , 于洋 . 数据增强的多模式时间感知序列推荐 [J ] . 电信科学 , 2024 , 40 ( 11 ): 66 - 78 .

Li J L , Wang R Q , Yu Y . Multi-pattern time-aware sequential recommendation with data augmentation [J ] . Telecommunications Science , 2024 , 40 ( 11 ): 66 - 78 .

Pankratz M , Lee H , LAD L , et al . GRU4RecBE: a hybrid session-based movie recommendation system (student abstract) [C ] // Proceedings of the Thirty-Sixth AAAI Conference on Artificial Intelligence, Thirty-Fourth Conference on Innovative Applications of Artificial Intelligence, and Twelfth Symposium on Educational Advances in Artificial Intelligence . New York : ACM Press , 2022 : 13029 - 13030 .

Tang J X , Wang K . Personalized top-N sequential recommendation via convolutional sequence embedding [C ] // Proceedings of the 11th ACM International Conference on Web Search and Data Mining . New York : ACM Press , 2018 : 565 - 573 .

王瑞琴 , 黄熠旻 , 纪其顺 , 等 . 注意力感知的边-节点交换图神经网络模型 [J ] . 电信科学 , 2024 , 40 ( 1 ): 106 - 114 .

Wang R Q , Huang Y M , Ji Q S , et al . Attention aware edge-node exchange graph neural network [J ] . Telecommunications Science , 2024 , 40 ( 1 ): 106 - 114 .

Kang W C , Mcauley J . Self-attentive sequential recommendation [C ] // Proceedings of the 2018 IEEE International Conference on Data Mining (ICDM) . Piscataway : IEEE Press , 2018 : 197 - 206 .

He X N , Deng K , Wang X , et al . LightGCN: simplifying and powering graph convolution network for recommendation [C ] // Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval . New York : ACM Press , 2020 : 639 - 648 .

He R N , Mcauley J . VBPR: visual Bayesian personalized ranking from implicit feedback [J ] . Proceedings of the AAAI Conference on Artificial Intelligence , 2016 , 30 ( 1 ): 144 - 150 .

Wei Y W , Wang X , Nie L Q , et al . MMGCN: multi-modal graph convolution network for personalized recommendation of micro-video [C ] // Proceedings of the 27th ACM International Conference on Multimedia . New York : ACM Press , 2019 : 1437 - 1445 .

Geng X , Li Y G , Wang L Y , et al . Spatiotemporal multi-graph convolution network for ride-hailing demand forecasting [C ] // Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence and Thirty-First Innovative Applications of Artificial Intelligence Conference and Ninth AAAI Symposium on Educational Advances in Artificial Intelligence . New York : ACM Press , 2019 : 3656 - 3663 .

Zhou X , Shen Z Q . A tale of two graphs: freezing and denoising graph structures for multimodal recommendation [C ] // Proceedings of the 31st ACM International Conference on Multimedia . New York : ACM Press , 2023 : 935 - 943 .

于洋 , 王瑞琴 . 基于对比增强时间感知自注意力机制的序列推荐 [J ] . 电信科学 , 2025 , 41 ( 1 ): 137 - 147 .

Yu Y , Wang R Q . Sequential recommendation based on contrast enhanced time-aware self-attention mechanism [J ] . Telecommunications Science , 2025 , 41 ( 1 ): 137 - 147 .

Xie X , Sun F , Liu Z Y , et al . Contrastive learning for sequential recommendation [C ] // Proceedings of the 2022 IEEE 38th International Conference on Data Engineering (ICDE) . Piscataway : IEEE Press , 2022 : 1259 - 1273 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于小波变换的RLS波束成形算法研究

小波分析法解决无线定位跟踪问题

基于动态有限域的保留格式加密算法

IRS辅助MIMO系统的低复杂度自适应相位优化算法

基于NOMA的双无人船海上通信资源分配与功率优化