面向用户多行为基于Transformer的个性化序列推荐方法和系统
By fusing multi-behavioral data and multi-scale temporal modeling of the Transformer model, combined with heterogeneous graphs and graph neural networks, the accuracy and diversity issues of existing personalized sequence recommendation systems under various types of shopping behaviors and noisy data are solved, achieving more efficient personalized recommendations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- INNER MONGOLIA UNIV OF TECH
- Filing Date
- 2024-08-30
- Publication Date
- 2026-07-17
AI Technical Summary
Existing personalized sequence recommendation systems are not very accurate when handling various types of shopping behaviors, cannot fully capture users' true interests, and are affected by noisy data, leading to the long-tail problem by ignoring the diversity of recommendation systems.
A multi-behavioral data fusion method is adopted, which combines the Transformer model for multi-scale time modeling and noise reduction. Heterogeneous graphs and graph neural networks are used to solve the long-tail problem. Graph contrastive learning is used to enhance the product embedding representation and fuse behavioral and product features.
It improves the accuracy and personalization of recommendations, better captures user interests and needs, enhances the efficiency and performance of the recommendation system, and provides fast and accurate recommendation services.
Smart Images
Figure CN119338543B_ABST