[1]王军锋,杨佳悦,李　钝.基于少数类加权和异常连通性的不平衡节点分类[J].郑州大学学报(工学版),2025,46(03):136-142.[doi:10.13705/j.issn.1671-6833.2024.06.019]
　WANG Junfeng,YANG Jiayue,LI Dun.Unbalanced Node Classification Based on Minority Class Weighted and Anomalous Connectivity[J].Journal of Zhengzhou University (Engineering Science),2025,46(03):136-142.[doi:10.13705/j.issn.1671-6833.2024.06.019]
点击复制

基于少数类加权和异常连通性的不平衡节点分类()
分享到：

《郑州大学学报(工学版)》[ISSN:1671-6833/CN:41-1339/T]

卷:
46
期数:
2025年03期
页码:
136-142
栏目:
出版日期:
2025-05-13

文章信息/Info

Title:
Unbalanced Node Classification Based on Minority Class Weighted and Anomalous Connectivity
文章编号:
1671-6833(2025)03-0136-07
作者:
王军锋;  杨佳悦;  李　钝
郑州大学 计算机与人工智能学院,河南 郑州 450001
Author(s):
WANG Junfeng;  YANG Jiayue;  LI Dun
School of Computer and Artificial Intelligence, Zhengzhou University, Zhengzhou 450001, China
关键词:
机器人账户;  类不平衡;  图结构;  少数类加权;  连通性
Keywords:
robot account;  class imbalance;  graph structure;  minority class weighted;  connectivity
分类号:
TP391
DOI:
10.13705/j.issn.1671-6833.2024.06.019
文献标志码:
A
摘要:
基于GNN的机器人检测方法在处理类不平衡问题时,忽略了少数类节点的重要性,同时未考虑图结构特有的链接性问题,使得节点分类效果不理想。针对现有方案的不足,提出了一种基于少数类加权和异常连通性裕度损失的类不平衡节点分类算法,将传统机器学习领域的不平衡分类思想扩展到图结构数据,在GraphSMOTE的基础上进行少数类加权聚合处理,以增强少数节点的特征聚合;在过采样阶段,利用SMOTE算法对不平衡数据进行处理,并考虑了节点表示和拓扑结构。同时,训练一个边缘生成器来建模关系信息,并引入异常连通性裕度损失,以提高GNN对链接异常性的感知,增强模型对连通性信息的学习。最后在公开的微博、Twitter虚假账户和BlogCatalog数据集上进行实验,与SMOTE、Re-weight、GraphSMOTE、DR-GCN和mGNN这5种方法的对比结果表明:所提算法平均ACC达到84.3%;在Kaggle数据集上,所提算法比mGNN模型准确度提升1.3%。
Abstract:
The robot detection methods based on GNN ignored the importance of minority class nodes when dealing with class imbalance problems, and did not consider the unique connectivity problem of graph structures, resulting in unsatisfactory node classification performance. Therefore, in response to the shortcomings of existing solutions, in this study, a class imbalanced node classification algorithm was proposed based on minority class weighted and abnormal connectivity margin loss, which extended the traditional imbalanced classification idea in the field of machine learning to graph structured data. Based on GraphSMOTE, minority class weighted aggregation was performed to enhance the feature aggregation of minority nodes. In the oversampling stage, the SMOTE algorithm was used to process imbalanced data, which considering node representation and topology structure. Simultaneously training an edge generator to model relational information and introducing anomalous connectivity margin loss to improve GNN′s perception of connectivity anomalies and enhance the model′s learning of connectivity information.Finally, experiments were conducted on publicly available Weibo, Twitter fake accounts, and BlogCatalog datasets. The comparison results with the five baselines of SMOTE, Re-weight, GraphSMOTE, DR-GCN, and mGNN showed that the average ACC of the algorithm proposed in this study reached 84.3%, with an accuracy improvement of 1.3% compared to the mGNN model on the Kaggle dataset.

参考文献/References:

[1]GUO Z W, WANG H. A deep graph neural networkbased mechanism for social recommendations[J]. IEEE Transactions on Industrial Informatics, 2021, 17(4): 2776-2783. 

[2]MOHAMMADREZAEI M, SHIRI M E, RAHMANI A M. Identifying fake accounts on social networks based ongraph analysis and classification algorithms[J]. Security and Communication Networks, 2018, 2018: 5923156. 
[3]田鸿朋, 张震, 张思源, 等. 复合可靠性分析下的不平衡数据证据分类[J]. 郑州大学学报(工学版), 2023, 44(4): 22-28. 
TIAN H P, ZHANG Z, ZHANG S Y, et al. Imbalanced data evidential classification with composite reliability [J]. Journal of Zhengzhou University (Engineering Science), 2023, 44(4): 22-28.
[4]PARK J, SONG J G, YANG E. GraphENS: neighbor-aware ego network synthesis for class-imbalanced node classification[EB/OL]. (2022-11-09)[2024-0104].https:∥specialsci.cn/detail/0425c6aa-3711-4e1fb070-d4e9bea2eb9b? resourceType=0. 
[5]ZHAO T X, ZHANG X, WANG S H. GraphSMOTE: imbalanced node classification on graphs with graph neural networks[C]∥Proceedings of the 14th ACM International Conference on Web Search and Data Mining.New York: ACM, 2021: 08826. 
[6]WANG K F, AN J, ZHOU M C, et al. Minority-weighted graph neural network for imbalanced node classification in social networks of Internet of people[J]. IEEE Internet of Things Journal, 2023, 10(1): 330-340. 
[7]SHI S H, QIAO K, CHEN C, et al. Over-sampling strategy in feature space for graphs based class-imbalanced bot detection[C]∥Companion Proceedings of the ACM on Web Conference 2024. New York: ACM, 2024: 06900. 
[8]CHEN D L, LIN Y K, ZHAO G X, et al. Topology-imbalance learning for semi-supervised node classification [EB/OL]. (2021-10-08)[2024-01-04]. http:∥arxiv.org/abs/2110.04099. 
[9]SONG J, PARK J, YANG E. TAM: topology-aware margin loss for class-imbalanced node classification[EB/ OL]. (2022-06-22)[2024-01-04]. http:∥arxiv. org/abs/2206.12917. 
[10] HAN H, WANG W Y, MAO B H. Borderline-SMOTE: a new over-sampling method in imbalanced data sets learning[M]∥Lecture Notes in Computer Science. Heidelberg: Springer Berlin Heidelberg, 2005: 878-887. 
[11] MATHEW J, LUO M, PANG C K, et al. Kernel-based SMOTE for SVM classification of imbalanced datasets[C] ∥IECON 2015 - 41st Annual Conference of the IEEE Industrial Electronics Society. Piscataway: IEEE, 2015: 1127-1132. 
[12]WANG K F, AN J, YU Z B, et al. Kernel local outlier factor-based fuzzy support vector machine for imbalanced classification[J]. Concurrency and Computation: Practice and Experience, 2021, 33(13): 1-10. 
[13] CHAWLA N V, LAZAREVIC A, HALL L O, et al. SMOTEBoost: improving prediction of the minority class in boosting[C]∥European Conference on Principles of Data Mining and Knowledge Discovery. Heidelberg: Springer, 2003: 107-119. 
[14] BANDINELLI N, BIANCHINI M, SCARSELLI F. Learning long-term dependencies using layered graph neural networks[C]∥The 2010 International Joint Conference on Neural Networks (IJCNN). Piscataway: IEEE, 2010: 1-8. 
[15] HAMILTON W L, YING R, LESKOVEC J, et al. Inductive representationlearning on large graphs[C]∥Advances in Neural Information Processing Systems. Lang Beach: NIPS, 2017:1025-1035. 
[16] DEFFERRARD M, BRESSON X, VANDERGHEYNST P. Convolutional neural networks on graphs with fast localized spectral filtering[C]∥Proceedings of the 30th International Conference on Neural Information Processing Systems. New York: ACM, 2016: 3844-3852. 
[17] KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[EB/OL]. (2017-0222)[2024-01-04]. http:∥arxiv.org/abs/1609.02907. 
[18] CHAWLA N V, BOWYER K W, HALL L O, et al. SMOTE: synthetic minority over-sampling technique[J]. Journal of Artificial Intelligence Research, 2002, 16: 321-357. 
[19] LIU L Q, LU Y, LUO Y, et al. Detecting " smart" spammers on social network: a topic model approach [EB/OL]. (2016-06-09)[2024-01-04]. http:∥arxiv.org/abs/1604.08504. 
[20] TANG L, LIU H. Relational learning via latent social dimensions[C]∥Proceedings of the 15th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2009: 817-826. 
[21] BO Y, MA X L. Sampling reweighting: boosting the performance of AdaBoost on imbalanced datasets[C]∥The 2012 International Joint Conference on Neural Networks (IJCNN). Piscataway: IEEE, 2012: 1-6. 
[22] SHI M, TANG Y F, ZHU X G, et al. Multi-class imbalancedgraph convolutional network learning[C]∥Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence. Yokohama: IJCAI, 2020: 28792885.

更新日期/Last Update: 2025-05-22