.png){kind=icon}
Return
DKGCN: Entity alignment based on dynamic graph inference in e-commerce platforms
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2026.133542.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In e-commerce platforms, given the dynamic characteristics of commodity entity information, temporal information is particularly important in entity alignment.We propose a dynamic graph neural network-based entity alignment framework tailored for e-commerce platforms, considering the dynamic changes of commodity data sources as a starting point. Firstly, addressing the dynamic structural changes in e-commerce platforms and utilizing the potential of Recurrent Neural Networks (RNNs) to extract intricate structural dependencies within K-core hierarchies, we propose a graph neural network named DKGCN based on K-core decomposition. DKGCN captures graph dynamics while maintaining local connectivity proximity and global structural similarity; furthermore, by passing messages between adjacent time steps, the model establishes consistency constraints. Secondly, to address the fusion of temporal and structural features, we construct a dynamic graph neural network-based structural-temporal embedding model.This model embeds structural and temporal information separately: the structural component focuses on network nodes and connectivity, while the temporal component records the dynamic evolution of nodes and edges across the timeline. These independent vectors are then fully integrated to obtain a comprehensive overall embedding vector. Finally, to tackle the data sparsity problem in e-commerce platforms, and considering the advantage of attention mechanisms in capturing implicit relations, we propose a temporal relational attention mechanism. This mechanism calculates the attention weights of each time point to reflect the importance of different moments in the time series for the alignment task. Experiments demonstrate that the proposed method effectively adapts to the evolving entity relations and attributes in e-commerce platforms, enhancing the accuracy and robustness of entity alignment.
Keywords:
Entity alignment
Dynamic graph neural networks
Temporal attention mechanism
K-core decomposition
E-commerce
Journal
IF:
7.5
Papers:
3.0W
Citations:
10.2W
Organization
No organization information available
Cited Papers
No cited papers available