{"title":"Weighted dynamic network link prediction based on graph autoencoder","authors":"Peng Mei, Yuhong Zhao, Jingyu Wang, Yefei Liang","doi":"10.1016/j.ins.2025.122507","DOIUrl":null,"url":null,"abstract":"<div><div>With the development of deep learning, Graph Autoencoders (GAE) within unsupervised learning frameworks have been widely applied to representation learning in dynamic networks. However, existing methods typically assume that the node set remains fixed across all time slices and ignore edge weight information, which limits the ability to capture network dynamics and distinguish the strength of node relationships. To address these issues, this paper proposes a weighted dynamic network link prediction framework based on GAE, called GAE_GGLA. This framework introduces an alignment module that can handle non-fixed node sets to adapt to dynamic network environments. Additionally, the edge weight matrix is used as a bias term in the graph attention network to calculate attention coefficients, guiding the learning of node features and enhancing their representational capacity. Furthermore, the GAE encoder employs graph convolution network (GCN) and long short-term memory (LSTM) networks to capture, respectively, structural features and temporal evolution. The alignment module connects different node sets through adjacent time slices, ensuring the continuity and consistency of network information. Finally, the GAE decoder reconstructs the adjacency matrix of the original graph to achieve link prediction. Experiments conducted on five different datasets demonstrate that the proposed method significantly outperforms state-of-the-art baselines.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"720 ","pages":"Article 122507"},"PeriodicalIF":6.8000,"publicationDate":"2025-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Information Sciences","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0020025525006395","RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"0","JCRName":"COMPUTER SCIENCE, INFORMATION SYSTEMS","Score":null,"Total":0}
引用次数: 0
Abstract
With the development of deep learning, Graph Autoencoders (GAE) within unsupervised learning frameworks have been widely applied to representation learning in dynamic networks. However, existing methods typically assume that the node set remains fixed across all time slices and ignore edge weight information, which limits the ability to capture network dynamics and distinguish the strength of node relationships. To address these issues, this paper proposes a weighted dynamic network link prediction framework based on GAE, called GAE_GGLA. This framework introduces an alignment module that can handle non-fixed node sets to adapt to dynamic network environments. Additionally, the edge weight matrix is used as a bias term in the graph attention network to calculate attention coefficients, guiding the learning of node features and enhancing their representational capacity. Furthermore, the GAE encoder employs graph convolution network (GCN) and long short-term memory (LSTM) networks to capture, respectively, structural features and temporal evolution. The alignment module connects different node sets through adjacent time slices, ensuring the continuity and consistency of network information. Finally, the GAE decoder reconstructs the adjacency matrix of the original graph to achieve link prediction. Experiments conducted on five different datasets demonstrate that the proposed method significantly outperforms state-of-the-art baselines.
期刊介绍:
Informatics and Computer Science Intelligent Systems Applications is an esteemed international journal that focuses on publishing original and creative research findings in the field of information sciences. We also feature a limited number of timely tutorial and surveying contributions.
Our journal aims to cater to a diverse audience, including researchers, developers, managers, strategic planners, graduate students, and anyone interested in staying up-to-date with cutting-edge research in information science, knowledge engineering, and intelligent systems. While readers are expected to share a common interest in information science, they come from varying backgrounds such as engineering, mathematics, statistics, physics, computer science, cell biology, molecular biology, management science, cognitive science, neurobiology, behavioral sciences, and biochemistry.