A method and system for temporal graph representation learning
By employing a multi-round iterative update temporal graph representation learning method, combined with Transformer and GCN models, the problem of insufficient temporal feature capture in temporal graphs is solved, thereby improving the accuracy of user behavior features and prediction accuracy.
CN115809347BActive Publication Date: 2026-06-16ALIPAY (HANGZHOU) INFORMATION TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ALIPAY (HANGZHOU) INFORMATION TECH CO LTD
- Filing Date
- 2022-12-14
- Publication Date
- 2026-06-16
Smart Images
Figure CN115809347B_ABST
Abstract
The embodiment of the specification discloses a kind of timing diagram representation learning method and system.The timing diagram includes node and edge, and the information of edge includes time information.The method includes: determining the subgraph containing target node and its neighbor node from timing diagram.Based on each edge in the subgraph, obtain node pair information;Wherein, node pair information includes the representation information of target node, the representation information of neighbor node on the edge, the information of the edge.For each of the neighbor node, sequentially process the node pair information containing the neighbor node by first network;Further obtain the updated representation information of target node and the updated representation information of each neighbor node.Process the updated representation information of target node and its neighbor node and the edge between target node and its neighbor node by second network, obtain the target representation information of target node.
Need to check novelty before this filing date? Find Prior Art