Graph Evidential Learning for Anomaly Detection

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wei, Chunyu, Hu, Wenji, Hao, Xingjia, Wang, Yunhai, Chen, Yueguo, Bai, Bing, Wang, Fei
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915316074807296
author Wei, Chunyu
Hu, Wenji
Hao, Xingjia
Wang, Yunhai
Chen, Yueguo
Bai, Bing
Wang, Fei
author_facet Wei, Chunyu
Hu, Wenji
Hao, Xingjia
Wang, Yunhai
Chen, Yueguo
Bai, Bing
Wang, Fei
contents Graph anomaly detection faces significant challenges due to the scarcity of reliable anomaly-labeled datasets, driving the development of unsupervised methods. Graph autoencoders (GAEs) have emerged as a dominant approach by reconstructing graph structures and node features while deriving anomaly scores from reconstruction errors. However, relying solely on reconstruction error for anomaly detection has limitations, as it increases the sensitivity to noise and overfitting. To address these issues, we propose Graph Evidential Learning (GEL), a probabilistic framework that redefines the reconstruction process through evidential learning. By modeling node features and graph topology using evidential distributions, GEL quantifies two types of uncertainty: graph uncertainty and reconstruction uncertainty, incorporating them into the anomaly scoring mechanism. Extensive experiments demonstrate that GEL achieves state-of-the-art performance while maintaining high robustness against noise and structural perturbations.
format Preprint
id arxiv_https___arxiv_org_abs_2506_00594
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Graph Evidential Learning for Anomaly Detection
Wei, Chunyu
Hu, Wenji
Hao, Xingjia
Wang, Yunhai
Chen, Yueguo
Bai, Bing
Wang, Fei
Machine Learning
Artificial Intelligence
Graph anomaly detection faces significant challenges due to the scarcity of reliable anomaly-labeled datasets, driving the development of unsupervised methods. Graph autoencoders (GAEs) have emerged as a dominant approach by reconstructing graph structures and node features while deriving anomaly scores from reconstruction errors. However, relying solely on reconstruction error for anomaly detection has limitations, as it increases the sensitivity to noise and overfitting. To address these issues, we propose Graph Evidential Learning (GEL), a probabilistic framework that redefines the reconstruction process through evidential learning. By modeling node features and graph topology using evidential distributions, GEL quantifies two types of uncertainty: graph uncertainty and reconstruction uncertainty, incorporating them into the anomaly scoring mechanism. Extensive experiments demonstrate that GEL achieves state-of-the-art performance while maintaining high robustness against noise and structural perturbations.
title Graph Evidential Learning for Anomaly Detection
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2506.00594