Graph Transformation Based on Heterogeneous Information Network for Graph Algorithms

In the finical risk control, the graph structure data increasingly show its unique charm, especially the heterogeneous information network (HIN). And, graph computing algorithms for the data mining based on network is the most popular way at the moment. But, most of them require the graph network mu...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of physics. Conference series Jg. 2575; H. 1; S. 12006 - 12011
Hauptverfasser: Liu, Zhenguo, Ma, Chao, Zhao, Jimiao, Hu, Haiyang, Yin, Xiongyi
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Bristol IOP Publishing 01.08.2023
Schlagworte:
Algorithms
Data mining
Data processing
Graph neural networks
Homogeneity
Neural networks
Physics
Risk management
ISSN:1742-6588, 1742-6596
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In the finical risk control, the graph structure data increasingly show its unique charm, especially the heterogeneous information network (HIN). And, graph computing algorithms for the data mining based on network is the most popular way at the moment. But, most of them require the graph network must have homogeneity and could not be applied to heterogeneous networks. Although the neural network models have some work in the HIN, which like heterogeneous graph neural networks, but these ways cannot provide enough interpretability due to its operations in black-box. In this paper, we summarize strategies to transform the HIN into homogeneous network. And we propose methods aim to rebuild HIN as several homogeneous networks at fine-grained level and greatly retain the original network topological structure information compared the previous which easy to lose sight of. The effectiveness of our method is verified by real data and community detection algorithms. In the experiment, the analysis found that our approached consistently perform promising results compared with the coarse-grained data processing on HIN.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1742-6588
1742-6596
DOI:10.1088/1742-6596/2575/1/012006