Parallel Discovery of Fixed-sized Connected k-Core Skyline Communities




Esmaeilian Ghahroudi, Parisa

Journal Title

Journal ISSN

Volume Title



Graphs are powerful when it comes to representing complex relationships between objects, where nodes and edges represent entities and relationships between them respectively. In recent years, the concept of community structures in graphs has gained significant attention due to its broad applications in various fields, such as social media analysis, physics, biology, and more. Community structures represent groups of nodes that have close relationships with each other, providing valuable insights into the underlying relationships within the graph. Graph nodes are often associated with attributes that contain valuable information, and it would be informative to take them into account when looking for communities. One way to do so is through the use of skyline communities, which represent community structures of a graph that are pareto optimal with respect to attribute values of nodes. In this study, we focus on k-Core subgraphs, where every node has a degree of at least k, and look for those holding skyline properties. We propose both sequential and parallel algorithms for discovering skyline k-Core subgraphs and perform experiments to investigate how input parameters, such as the dataset, the size of the community, the number of attribute dimensions, etc., affect the performance of our solution. Our proposed approach is a progressive algorithm that can be stopped at any point, providing the assurance that any output obtained is a skyline community. We demonstrate the effectiveness of our approach on a large dataset which is able to achieve acceleration rates as high as 10x over the state-of-the-art method. Moreover, the parallelised version attains super-linear acceleration rates with 2-3 cores (2.25x and 3.29x respectively) and a speedup as high as 34x over the sequential version when utilising 48 cores.



skyline, community discovery, k-core, parallel computing