KGNIE: A Learning Method for Estimating Node Importance in Knowledge Graphs

Dec 17, 2023·
Yin Chen
Jianke Yu
Jianke Yu
,
Qing Sima
,
Jinghao Wang
,
Yanping Wu
,
Xiaoyang Wang
· 1 min read
CCF-C
Abstract
Estimating node importance is critical in graph data mining, benefiting various downstream applications such as social network analysis and recommendation systems. Existing approaches face challenges when dealing with complex knowledge graphs due to the abundance of predicate and entity information. This paper introduces KGNIE, an efficient knowledge graph node importance estimation network. KGNIE considers the rich predicate attributes and entity types, and utilizes local and global information encoders to generate node embeddings with different importance-related information. An attention-based fusion module is employed to balance the two encoders. A node importance decoder is proposed to map node embed dings to importance scores based on entity types. Furthermore, we introduce a margin ranking loss to determine relative node importance rankings. We conducted extensive experiments on real-world knowledge graphs, demonstrating that our model outperforms existing approaches across all evaluation metrics.
Type
Publication
IEEE International Conference on High Performance Computing and Communications
Status
Peer-reviewed
Awards
CCF-C
IEEE International Conference on High Performance Computing and Communications · 2023
publications

IEEE International Conference on High Performance Computing and Communications, 2023

Estimating node importance is critical in graph data mining, benefiting various downstream applications such as social network analysis and recommendation systems. Existing approaches face challenges when dealing with complex knowledge graphs due to the abundance of predicate and entity information. This paper introduces KGNIE, an efficient knowledge graph node importance estimation network. KGNIE considers the rich predicate attributes and entity types, and utilizes local and global information encoders to generate node embeddings with different importance-related information. An attention-based fusion module is employed to balance the two encoders. A node importance decoder is proposed to map node embed dings to importance scores based on entity types. Furthermore, we introduce a margin ranking loss to determine relative node importance rankings. We conducted extensive experiments on real-world knowledge graphs, demonstrating that our model outperforms existing approaches across all evaluation metrics.

Jianke Yu
Authors
Jianke Yu (he/him)
PhD (2026) in Graph Machine Learning & Databases
I completed my PhD at the University of Technology Sydney (UTS) in 2026, supervised by Prof. Ying Zhang, A/Prof. Lu Qin and Dr Hanchen Wang. My research develops machine learning algorithms for graph-structured data and database systems — graph neural network architectures, and learning-based methods that improve database and data-mining algorithms (fraud detection, graph similarity, supergraph search, malware detection). My work has appeared in KDD and IEEE TKDE. I am joining Zhejiang Gongshang University as a faculty member in 2027.