A two-phase streaming edge partitioning algorithm for large-scale uncertain graphs

Uncertain graphs, which model uncertainty in edges, are used in various domains such as social networks, biological networks, and recommendation systems. As the scale of these graphs continues to grow, efficient distributed analysis becomes crucial, hinging on effective graph partitioning. However,...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Cluster computing Ročník 28; číslo 11; s. 746
Hlavní autoři: Cui, Huanqing, Chang, Anfu, Hu, Kekun, Dong, Gang, Zhu, Jinbin
Médium: Journal Article
Jazyk:angličtina
Vydáno: New York Springer US 01.10.2025
Springer Nature B.V
Témata:
Algorithms
Apexes
Clustering
Computer Communication Networks
Computer Science
Graph theory
Graphs
Load balancing
Operating Systems
Optimization
Partitioning
Partitions (mathematics)
Probability
Probability distribution
Processor Architectures
Recommender systems
Skewness
Social networks
Uncertainty
Graph partitioning
Distributed graph computing
Uncertain graph
Degree advantage
ISSN:1386-7857, 1573-7543
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Uncertain graphs, which model uncertainty in edges, are used in various domains such as social networks, biological networks, and recommendation systems. As the scale of these graphs continues to grow, efficient distributed analysis becomes crucial, hinging on effective graph partitioning. However, existing partitioning algorithms are primarily designed for deterministic graphs, which do not account for edge uncertainties. When applied to uncertain graphs, these methods often result in too many replicated vertices and imbalanced partition loads. This paper presents a streaming edge partitioning algorithm TPEP (Two-Phase Edge Partitioning) to partition the large-scale uncertain graphs. Firstly, TPEP formulates the partitioning problem as an optimization problem of minimizing the number of replicated vertices, using the expected number of edges to evaluate partition load. Then, TPEP uses a score function to guide the streaming partitioning process, where the score function consists of the ratio of replicated vertices, load balance, and the number of vertices within the same cluster. To compare the degrees of two vertices, the concept of degree advantage is introduced. Experimental results indicate that TPEP outperforms existing algorithms by achieving an average reduction of 34.3% in replicated vertices, an average reduction of 33.1% in load balance, and the optimal edge probability distribution.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1386-7857
1573-7543
DOI:10.1007/s10586-025-05338-5