Article
Computer Science, Artificial Intelligence
Chun Wang, Shirui Pan, Celina P. Yu, Ruiqi Hu, Guodong Long, Chengqi Zhang
Summary: Node clustering is a method to partition the vertices in a graph into multiple groups or communities. Existing studies mainly focus on developing deep learning approaches to learn a latent representation of nodes, followed by simple clustering methods like k-means. In this paper, a clustering-directed deep learning approach called DNENC is proposed to encode the topological structure and node content in a graph, and generate soft labels for a self-training process to iteratively refine the node clustering results.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Information Systems
Zhe Chen, Aixin Sun, Xiaokui Xiao
Summary: The article introduces an incremental community detection framework inc-AGGMMR, which maps attributes into the network by constructing an augmented graph and balances the contribution between attribute and topological information through weight adjustment.
ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA
(2021)
Article
Computer Science, Information Systems
Shunjie Yuan, Hefeng Zeng, Ziyang Zuo, Chao Wang
Summary: This study proposes an overlapping community detection method CDMG based on Graph Convolutional Networks (GCN) to maximize the Markov stability of community structure. Extensive experiments demonstrate the superiority of CDMG compared to other established community detection algorithms. The performance of CDMG can be further improved by utilizing the optimal Markov time, which is found using a trichotomy-based method based on the influence of Markov time on CDMG performance.
COMPUTER COMMUNICATIONS
(2023)
Article
Computer Science, Information Systems
G. Victor Daniel, Kandasamy Chandrasekaran, Venkatesan Meenakshi, Prabhavathy Paneer
Summary: This paper presents a robust graph neural network model called AnomEn for anomaly detection. The model addresses the issues of false positives and false negatives caused by aggregate operations in graph neural networks. It employs a weighted aggregate mechanism to emphasize a node's own features and preserve their original characteristics. Experimental results demonstrate the effectiveness of the proposed method, which outperforms existing methods in both node and edge anomaly detection tasks.
Article
Computer Science, Artificial Intelligence
Hongtao Liu, Jiahao Wei, Tianyi Xu
Summary: In this paper, a new community detection method called CPGC is proposed, which combines the community perspective and graph convolution network to address the challenges of overlapping communities in attributed networks. CPGC achieves state-of-the-art results in nonoverlapping or overlapping communities, as demonstrated by experiments on various real-world networks.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Chaobo He, Yulong Zheng, Junwei Cheng, Yong Tang, Guohua Chen, Hai Liu
Summary: This paper proposes a semi-supervised overlapping community detection method named SSGCAE, which is based on graph neural networks. It addresses the problems of link and attribute fusion, prior information integration, and overlapping community detection in attributed graphs.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Haiping Ma, Zhenjie Liu, Xingyi Zhang, Lei Zhang, Hao Jiang
Summary: The paper proposed a dual-population-based multi-objective evolutionary algorithm, DP-MOEA, for community detection in attributed networks, aiming to balance topology structure and node attribute. Experimental results show the superior performance of the proposed DP-MOEA over eight state-of-art evolutionary algorithms, especially when dealing with unclear community structure or heterogeneous node attributes within one community.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Xinchuang Zhou, Lingtao Su, Xiangju Li, Zhongying Zhao, Chao Li
Summary: This paper proposes a community detection algorithm based on unsupervised attributed network embedding (CDBNE) to address the challenges in community detection tasks. By simultaneously learning network structure, attribute information, and clustering-oriented representation, CDBNE outperforms state-of-the-art methods, as demonstrated by experiments.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Jingcan Duan, Bin Xiao, Siwei Wang, Haifang Zhou, Xinwang Liu
Summary: Recently, there has been increasing attention on graph anomaly detection on attributed networks in the data mining and machine learning communities. This detection aims not only at attribute anomalies, but also at suspicious topological-abnormal nodes that exhibit collective anomalous behavior. However, existing methods overlook the importance of recognizing collective patterns for improving topology anomaly detection performance. In this study, a new framework called ARISE is proposed for graph anomaly detection on attributed networks via substructure awareness. ARISE focuses on identifying abnormalities in the substructures of the graph. It establishes a region proposal module to discover high-density substructures as suspicious regions and utilizes average node-pair similarity to measure the topology anomaly degree. Additionally, a graph contrastive learning scheme is introduced to extract better embeddings of node attributes and observe attribute anomalies. Experimental results on benchmark datasets demonstrate that ARISE significantly improves detection performance compared to state-of-the-art algorithms for attributed networks anomaly detection (ANAD).
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Information Systems
Kamal Berahmand, Sogol Haghani, Mehrdad Rostami, Yuefeng Li
Summary: The diffusion method is a major approach for community detection in complex networks. The LPA algorithm, which mimics epidemic contagion, is efficient but has some issues. This paper proposes a new version of the LPA algorithm for attributed graphs that solves problems related to instability and low quality, and improves node selection and updating mechanisms. Experimental results show that the proposed method outperforms other state-of-the-art attributed graph clustering methods in terms of efficiency and accuracy.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Xiaofeng Wang, Jianhua Li, Li Yang, Hongmei Mi
Summary: Community detection in attributed networks has become a challenging problem in network science, and recent advancements in network representation learning, particularly those based on graph convolutional networks (GCN), have attracted considerable attention. The proposed method in this study integrates a label sampling model and GCN into an unsupervised learning framework to uncover underlying community structures in attributed networks. Experimental results show that the proposed method outperforms current state-of-the-art community detection algorithms.
Article
Computer Science, Artificial Intelligence
Marco De Luca, Anna Rita Fasolino, Antonino Ferraro, Vincenzo Moscato, Giancarlo Sperli, Porfirio Tramontana
Summary: In this paper, a novel heterogeneous graph-based model is proposed to capture and handle the complex and strongly-correlated information of a software Developer Social Network (DSN) for analytic tasks. The problem of automatically discovering communities of software developers sharing interests for similar projects is addressed using Social Network Analysis (SNA) findings, and graph embedding techniques are utilized to overcome the large graph size. The proposed approach is evaluated against state-of-the-art approaches in terms of efficiency and effectiveness using the GitHub dataset.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Engineering, Multidisciplinary
Kamal Berahmand, Mehrnoush Mohammadi, Farid Saberi-Movahed, Yuefeng Li, Yue Xu
Summary: Community detection is an important research topic in machine learning, but most existing methods only consider the network's topology structure, neglecting the advantage of using node attribute information. To solve this problem, we propose a novel Augment Graph Regularization Nonnegative Matrix Factorization method, which performs unexpectedly well in attributed networks.
IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Lun Hu, Xiangyu Pan, Hong Yan, Pengwei Hu, Tiantian He
Summary: Community detection is a fundamental task in cluster analysis, crucial for understanding complex network systems. With the increase in richness and variety of attribute information, detecting communities in attributed graphs has become more challenging.
INTEGRATED COMPUTER-AIDED ENGINEERING
(2021)
Article
Chemistry, Multidisciplinary
Hairu Luo, Peng Jia, Anmin Zhou, Yuying Liu, Ziheng He
Summary: A deep learning-based framework named BND is proposed in this paper for quickly and accurately detecting bridge nodes in networks. By considering the multi-dimensional attributes and structural characteristics of nodes, an attribute graph is constructed to extract bridging-related attributes. Graph neural network layers are used to process the attribute graph in the deep learning model, and fully connected layers are added to improve the classification effect. The experiments show that the framework can effectively capture network topology information and accurately detect bridge nodes.
APPLIED SCIENCES-BASEL
(2022)
Article
Mathematics, Applied
Petr Chunaev
JOURNAL OF MATHEMATICAL INEQUALITIES
(2015)
Article
Mathematics
Petr Chunaev, Vladimir Danchenko
JOURNAL OF APPROXIMATION THEORY
(2016)
Article
Mathematics
Petr Chunaev
JOURNAL OF GEOMETRIC ANALYSIS
(2017)
Article
Mathematics
Petr Chunaev, Vladimir Danchenko
JOURNAL OF APPROXIMATION THEORY
(2018)
Article
Mathematics
Petr Chunaev, Joan Mateu, Xavier Tolsa
MATHEMATISCHE ZEITSCHRIFT
(2020)
Article
Mathematics
Petr Chunaev, Joan Mateu, Xavier Tolsa
JOURNAL D ANALYSE MATHEMATIQUE
(2019)
Article
Mathematics, Applied
Petr Chunaev, Vladimir Danchenko
COMPLEX ANALYSIS AND OPERATOR THEORY
(2020)
Article
Mathematics
Petr Chunaev
JOURNAL OF APPROXIMATION THEORY
(2020)
Article
Computer Science, Information Systems
Petr Chunaev, Timofey Gradov, Klavdiya Bochenina
Summary: The weight-based fusion model (WBFM) is a simple and efficient model for modularity-driven community detection in node-attributed social networks. This paper reveals the mathematical machinery of the WBFM, proposes a non-manual parameter tuning scheme, and develops a tunable Leiden-based ASN CD algorithm.
SOCIAL NETWORK ANALYSIS AND MINING
(2021)
Article
Engineering, Electrical & Electronic
Elizaveta Stavinova, Ilyas Varshavskiy, Petr Chunaev, Ivan Derevitskii, Alexander Boukhanovsky
Summary: Dynamic pricing is widely used in various industries, but it has complex effects on human behavior. This paper proposes a data-driven method, DPRank, to evaluate and compare dynamic pricing systems. By comparing hidden and exposed models using a Monte Carlo simulation, the quality difference between the two models can vary significantly, indicating potential for improvement.
Proceedings Paper
Computer Science, Interdisciplinary Applications
Elizaveta Stavinova, Anastasiia Timoshina, Petr Chunaev
Summary: The study developed an open Internet data-based approach for forecasting the volume of mortgage loans in the residential real estate market of Saint Petersburg, Russia. By utilizing predictor variables like Yandex search queries, Russian Central Bank rates, and dollar/ruble exchange rates in an ARIMAX model, the forecasting quality was improved.
10TH INTERNATIONAL YOUNG SCIENTISTS CONFERENCE IN COMPUTATIONAL SCIENCE (YSC2021)
(2021)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Elizaveta Stavinova, Petr Chunaev, Klavdiya Bochenina
Summary: Dynamic pricing is a modern tool used by railways, airlines, and bus companies to increase revenue by adjusting ticket prices according to passenger demand. For passengers, deciding when to buy tickets in order to save money is crucial, and forecasting ticket prices typically involves using historical price data and departure features as predictor variables. The study investigates whether search engine query open data can enhance ticket price forecast quality for statistical and artificial neural network-based models, and the experiments show that this is indeed the case for railway ticket dynamic prices.
10TH INTERNATIONAL YOUNG SCIENTISTS CONFERENCE IN COMPUTATIONAL SCIENCE (YSC2021)
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Petr Chunaev, Timofey Gradov, Klavdiya Bochenina
9TH INTERNATIONAL YOUNG SCIENTISTS CONFERENCE IN COMPUTATIONAL SCIENCE, YSC2020
(2020)
Article
Mathematics
V. I. Danchenko, M. A. Komarov, P. V. Chunaev
RUSSIAN MATHEMATICS
(2018)