Community detection with Greedy Modularity disassembly strategy

Heru Cahya Rustamaji,Wisnu Ananta Kusuma,Sri Nurdiati,Irmanida Batubara

DOI: https://doi.org/10.1038/s41598-024-55190-7

IF: 4.6

2024-02-27

Scientific Reports

Abstract:Community detection recognizes groups of densely connected nodes across networks, one of the fundamental procedures in network analysis. This research boosts the standard but locally optimized Greedy Modularity algorithm for community detection. We introduce innovative exploration techniques that include a variety of node and community disassembly strategies. These strategies include methods like non-triad creating, feeble, random as well as inadequate embeddedness for nodes, as well as low internal edge density, low triad participation ratio, weak, low conductance as well as random tactics for communities. We present a methodology that showcases the improvement in modularity across the wide variety of real-world and synthetic networks over the standard approaches. A detailed comparison against other well-known community detection algorithms further illustrates the better performance of our improved method. This study not only optimizes the process of community detection but also broadens the scope for a more nuanced and effective network analysis that may pave the way for more insights as to the dynamism and structures of its functioning by effectively addressing and overcoming the limitations that are inherently attached with the existing community detection algorithms.

multidisciplinary sciences

What problem does this paper attempt to address?

The paper aims to address the local optimum problem in community detection and improve the quality of community detection by enhancing the Greedy Modularity algorithm. Specifically, the paper introduces innovative exploration techniques, including node and community dismantling strategies such as non-triangle creation, weak connections, random and low embeddedness nodes, as well as low internal edge density, low triangle participation ratio, weak connections, low conductance, and random community dismantling strategies. These methods demonstrate modularity improvements over standard methods in various real-world and synthetic networks and provide detailed comparisons with other well-known community detection algorithms, further proving the superior performance of the improved methods.

Moreover, the research not only optimizes the community detection process but also broadens the scope of more detailed and effective network analysis, providing new ways to understand and overcome the inherent limitations of existing community detection algorithms. By avoiding local optima, this approach can detect more accurate and meaningful community structures in complex networks, thereby having a significant impact on the field of network analysis and pointing the way for future research and applications.