Academic anomaly citation group detection based on local extended community detection

doi:10.11772/j.issn.1001-9081.2023050702

Abstract

Abstract:

Some scholars in the academic social network may form anomaly citation groups， and excessively cite each other’s papers for profit. Most of the existing anomaly group detection algorithms separate community detection from node representation learning， which leads to the limited performance of anomaly group detection. To deal with the issue， a Group Anomaly Detection based on Local extended community detection （GADL） algorithm was proposed. The author anomaly citation features were extracted by using semantic information such as research field and title content of the paper. An extension metric function based on node transition similarity， node community membership， citation anomaly and BFS （Breath First Search） depth was defined. The optimal anomaly detection performance could be obtained by combining anomaly community detection and anomaly node detection， and jointly optimizing them in a unified framework. Compared with ALP algorithm， the proposed algorithm improved the Area Under Curve （AUC） by 6.07%， 5.35% and 3.38% respectively on the ACM， DBLP1， and DBLP2 datasets.Experimental results on real datasets show that GADL can effectively detect academic anomaly citations.

Key words: academic social network, graph anomaly detection, academic anomaly citation, Graph Neural Network (GNN), local extended community detection

摘要：

学术社交网络中的某些学者可能组成异常引用群体，相互之间过度引用彼此的文章以谋取利益。现有的异常群体检测算法大多将社区检测与节点表示学习分离，导致最终异常群体检测性能受限。为此，提出一种基于局部扩展社区发现的异常引用群体检测（GADL）算法。所提算法利用论文研究领域、标题内容等语义信息提取作者异常引用特征；定义基于节点转移相似度、节点社区隶属度、引用异常度和广度优先遍历（BFS）深度的扩展度量函数；结合异常社区发现和异常节点检测，在统一框架下对二者联合优化，可获得最优的异常检测性能。在ACM、DBLP1和DBLP2数据集上，相较于ALP算法，所提算法分别提高了6.07%、5.35%和3.38%。在真实数据集上的实验结果表明，所提算法可有效地检测异常学术引用。

关键词: 学术社交网络, 图异常检测, 学术异常引用, 图神经网络, 局部扩展社区发现

CLC Number:

TP391.4

Xinrui LIN, Xiaofei WANG, Yan ZHU. Academic anomaly citation group detection based on local extended community detection[J]. Journal of Computer Applications, 2024, 44(6): 1855-1861.

林欣蕊, 王晓菲, 朱焱. 基于局部扩展社区发现的学术异常引用群体检测[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1855-1861.

Figures/Tables 8

References 18

1	FISTER I， Jr， FISTER I， PERC M. Toward the discovery of citation cartels in citation networks［J］. Frontiers in Physics， 2016， 4： 49.
2	BAI X， XIA F， LEE I， et al. Identifying anomalous citations for objective evaluation of scholarly article impact［J］. PLoS ONE， 2016， 11（9）： e0162364.
3	TOSYALI A， KIM J， CHOI J， et al. New node anomaly detection algorithm based on nonnegative matrix factorization for directed citation networks［J］. Annals of Operations Research， 2020， 288： 457-474.
4	蔡丽伟. 学术网络中异常引用的识别［D］. 大连：大连理工大学， 2020： 11-46.
	CAI L W. Anomalous citation detection in academic networks ［D］. Dalian： Dalian University of Technology， 2020：11-46.
5	LIU J， XIA F， FENG X， et al. Deep graph learning for anomalous citation detection［J］. IEEE Transactions on Neural Networks and Learning Systems， 2022， 33（6）：2543-2557.
6	HUANG L， ZHU Y， GAO Y， et al. Hybrid-order anomaly detection on attributed networks［J］. IEEE Transactions on Knowledge and Data Engineering， 2023， 35（12）： 12249-12263.
7	HU S， SHAO M. Dual perspective contrastive learning based subgraph anomaly detection on attributed networks［C］// Proceedings of the 31st International Conference on Artificial Neural Networks. Berlin： Springer， 2022： 481-493.
8	HUANG P， HUANG Y， WANG W， et al. Deep embedding network for clustering［C］// Proceedings of the 2014 22nd International Conference on Pattern Recognition. Piscataway： IEEE， 2014：1532-1537.
9	TIAN F， GAO B， CUI Q， et al. Learning deep representations for graph clustering［C］// Proceedings of the 28th AAAI Conference on Artificial Intelligence. Menlo Park： AAAI Press， 2014： 1293-1299.
10	王晓菲，朱焱. 基于深度强化学习的异常学术引用检测［J］. 计算机工程与设计， 2023， inpress.
	WANG X F， ZHU Y. Anomaly detection of academic citations based on deep reinforcement learning［J］. Computer Engineering and Design， 2023， inpress
11	BLONDEL V D， J-L GUILLAUME， LAMBIOTTE R， et al. Fast unfolding of communities in large networks ［EB/OL］.［2023-03-01］. .
12	SHANG R， ZHANG W， ZHANG J， et al. Local community detection based on higher-order structure and edge information［J］. Physica A： Statistical Mechanics and its Applications， 2022， 587： 126513.
13	PURWITASARI D， ILMI A B， FATICHAH C， et al. Conflict of interest based features for expert classification in bibliographic network［C］// Proceedings of the 2018 International Conference on Computer Engineering， Network and Intelligent Multimedia. Piscataway： IEEE， 2018： 54-59.
14	ILMI A B， PURWITASARI D， FATICHAH C. Siamese long short-term memory for detecting conflict of interest on scientific papers［J］. IPTEK The Journal for Technology and Science， 2019， 30（2）： 49-51.
15	LI Z， CHEN X， SONG J， et al. Adaptive label propagation for group anomaly detection in large-scale networks［J］. IEEE Transactions on Knowledge and Data Engineering， 2023， 35（12）： 12053-12067.
16	DING X， ZHANG J， YANG J. A robust two-stage algorithm for local community detection［J］. Knowledge-Based Systems， 2018， 152： 188-199.
17	YIN H， BENSON A R， LESKOVEC J， et al. Local higher-order graph clustering［C］// Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York： ACM， 2017： 555-564.
18	GUO K， HUANG X， WU L， et al. Local community detection algorithm based on local modularity density［J］. Applied Intelligence， 2022， 52： 1238-1253.

数据集	节点数	边数	异常占比	年份
ACM	23 498	46 018	0.072	2011—2013
DBLP1	74 680	186 050	0.064	2007—2009
DBLP2	143 987	366 079	0.061	2009—2010

数据集	节点数	边数	异常占比	年份
ACM	23 498	46 018	0.072	2011—2013
DBLP1	74 680	186 050	0.064	2007—2009
DBLP2	143 987	366 079	0.061	2009—2010

算法	ACM		DBLP1		DBLP2
算法	AUC	Macro-F1	AUC	Macro-F1	AUC	Macro-F1
GCoLA	0.687 8	0.635 7	0.672 8	0.605 0	0.680 4	0.626 1
ALP	0.915 6	0.702 5	0.936 2	0.734 1	0.954 8	0.786 3
GADL	0.9712	0.7469	0.9863	0.8285	0.9871	0.8794

算法	ACM		DBLP1		DBLP2
算法	AUC	Macro-F1	AUC	Macro-F1	AUC	Macro-F1
GCoLA	0.687 8	0.635 7	0.672 8	0.605 0	0.680 4	0.626 1
ALP	0.915 6	0.702 5	0.936 2	0.734 1	0.954 8	0.786 3
GADL	0.9712	0.7469	0.9863	0.8285	0.9871	0.8794

算法	ACM		DBLP1		DBLP2
算法	AUC	Macro-F1	AUC	Macro-F1	AUC	Macro-F1
RTLCD	0.807 8	0.535 7	0.842 8	0.625 0	0.854 4	0.656 1
MAPPR	0.874 2	0.698 4	0.906 4	0.779 7	0.918 3	0.797 8
LCDMD	0.939 8	0.710 4	0.949 0	0.801 4	0.955 3	0.819 0
HSEI	0.960 2	0.735 2	0.972 5	0.820 6	0.976 1	0.863 5
GADL	0.9712	0.7469	0.9863	0.8285	0.9871	0.8794