Page 67 - 《软件学报》2026年第4期

1508 软件学报 2026 年第 37 卷第 4 期

总体而言, 在所有 DIG 比率下的实验结果均表明, EXDIG 在不同程度的缺失图组件情况下仍能保持可解释
性. 模型在不同扰动级别下始终能够稳定地识别关键节点和边, 进一步验证了其在动态不完整图环境下的鲁棒性
和可靠性.
5 总 结

我们提出了 EXDIG 框架, 以解决动态不完整图 (DIG) 中的异常检测问题, 通过应对特征和结构的不完整性来
提升模型性能. 大量实验验证了该方法在不同动态不完整条件下的鲁棒性和有效性, 能够恢复 GNN 的表示能力,
并适用于多种下游任务. 此外, EXDIG 通过识别关键组成部分, 提高了异常检测结果的可解释性. 该框架为扩展到
更复杂、更贴近现实的工业生产图场景提供了新的可能性. 未来的研究将重点优化 EXDIG 的计算效率和可扩展
性, 使其能够应用于包含数百万个节点和边的大规模图数据. 具体而言, 一是探索针对演化图流的动态图数据流建
模框架, 以实时捕捉动态不完整图中结构与特征的快速演化; 二是研究轻量化的结构增强掩码策略, 以降低模型的
计算负担, 并提升在大规模动态图场景下的适应性. 为此, 我们将探索优化模型架构的方法, 并结合分布式或并行
计算策略, 以提升其在真实大规模部署中的可行性.

References
[1] Ma XX, Wu J, Xue S, Yang J, Zhou C, Sheng QZ, Xiong H, Akoglu L. A comprehensive survey on graph anomaly detection with deep
learning. IEEE Trans. on Knowledge and Data Engineering, 2023, 35(12): 12012–12038. [doi: 10.1109/tkde.2021.3118815]
[2] Kim H, Lee BS, Shin WY, Lim S. Graph anomaly detection with graph neural networks: Current status and challenges. IEEE Access,
2022, 10: 111820–111829. [doi: 10.1109/ACCESS.2022.3211306]
[3] Noble CC, Cook DJ. Graph-based anomaly detection. In: Proc. of the 9th ACM SIGKDD Int’l Conf. on Knowledge Discovery and Data
Mining. Washington: ACM, 2003. 631–636. [doi: 10.1145/956750.956831]
[4] Kipf TN, Welling M. Semi-supervised classification with graph convolutional networks. arXiv:1609.02907, 2016.
[5] Veličković P, Cucurull G, Casanova A, Romero A, Liò P, Bengio Y. Graph attention networks. In: Proc. of the 6th Int’l Conf. on
Learning Representations. Vancouver: OpenReview.net, 2018. [doi: 10.48550/arXiv.1710.10903]
[6] Hamilton WL, Ying R, Leskovec J. Inductive representation learning on large graphs. In: Proc. of the 31st Int’l Conf. on Neural
Information Processing Systems. Red Hook: Curran Associates Inc., 2017. 1025–1035. [doi: 10.5555/3294771.3294869]
[7] Song LY, Ma ZY, Li ZH, Shang XQ. Review on temporal graph neural networks for financial risk prediction. Ruan Jian Xue Bao/Journal
of Software, 2024, 35(8): 3897–3922 (in Chinese with English abstract). http://www.jos.org.cn/1000-9825/7087.htm [doi: 10.13328/j.cnki.
jos.007087]
[8] Yu H, Liu ZY, Luo XF. Barely supervised learning for graph-based fraud detection. In: Proc. of the 38th AAAI Conf. on Artificial Int’l
and 36th Conf. on Innovative Applications of Artificial Intelligence and 14th Symp. on Educational Advances in Artificial Intelligence.
Menlo Park: AAAI Press, 2024. 16548–16557. [doi: 10.1609/aaai.v38i15.29593]
[9] Liu ZY, Yu H, Luo XF. Federated graph anomaly detection via disentangled representation learning. In: Proc. of the 2025 ACM on Web
Conf. Sydney: ACM, 2025. 1216–1224. [doi: 10.1145/3696410.3714567]
[10] Wang RJ, Mou S, Wang X, Xiao WP, Ju Q, Shi C, Xie X. Graph structure estimation neural networks. In: Proc. of the 2021 Web Conf.
Ljubljana: ACM, 2021. 342–353. [doi: 10.1145/3442381.3449952]
[11] Lai XC, Zhang Z, Chen H, Zhang LY, Li ZH, Lu W. Tracking-removed neural network with graph information for classification of
incomplete data. Applied Intelligence, 2025, 55(4): 256. [doi: 10.1007/s10489-024-06031-7]
[12] Zhou BS, Ji JT, Gu ZB, Zhou ZH, Ding GY, Feng SH. One-step graph-based incomplete multi-view clustering. Multimedia Systems,
2024, 30(1): 32. [doi: 10.1007/s00530-023-01225-4]
[13] Zhang H, Chen XH, Zhang EH, Wang LP. Incomplete multi-view learning via consensus graph completion. Neural Processing Letters,
2023, 55(4): 3923–3952. [doi: 10.1007/s11063-022-10973-9]
[14] Ge Y, Chen SC. Graph convolutional network for recommender systems. Ruan Jian Xue Bao/Journal of Software, 2020, 31(4):
1101–1112 (in Chinese with English abstract). http://www.jos.org.cn/1000-9825/5928.htm [doi: 10.13328/j.cnki.jos.005928]
[15] Liu J, Shang XQ, Song LY, Tan YC. Progress of graph neural networks on complex graph mining. Ruan Jian Xue Bao/Journal of
Software, 2022, 33(10): 3582–3618 (in Chinese with English abstract). http://www.jos.org.cn/1000-9825/6626.htm [doi: 10.13328/j.cnki.
jos.006626]